# sourcesignal documentation > How to set up sourcesignal, track the questions buyers ask AI engines, work the opportunity queue, run outreach from your own inbox and report what changed. Every article of https://www.sourcesignal.ai/docs in reading order, 73 in all. Each begins with its title, its section, the day it was updated and the page's address. The index with one line per article: https://www.sourcesignal.ai/llms.txt. An agent connecting over MCP reads https://www.sourcesignal.ai/outreach-for-agents/brief.md as well. The documentation describes what a customer sees and can do, never how a verdict is computed: how a page earns its priority, how a mention is read or how a page type is chosen is not published. --- # What sourcesignal does > sourcesignal tracks the questions buyers ask AI engines, shows you the pages those answers cite, and helps you work your brand onto them. sourcesignal documentation, Getting started. Updated 2026-10-20. The page for people: https://www.sourcesignal.ai/docs/getting-started/what-sourcesignal-does. Every article: https://www.sourcesignal.ai/llms.txt Buyers ask AI engines for recommendations and shortlists. Those answers lean on third-party pages: listicles, comparisons, reviews, directories and forum threads, among others. sourcesignal tracks the questions that matter to your brand, reads the answers across the engines you choose, and turns each page those answers cite into an opportunity you can act on. You decide what to pursue. The platform does the collecting, grading, drafting and measuring. ## How the loop works 1. **You define what visibility means.** The questions (prompts) buyers ask, the competitors you measure against and the markets you sell in. 2. **Answers are collected.** The engines your project tracks answer those questions on a schedule. Each answer is kept in full, so you can see who was named, whether your brand was recommended and which pages were cited. 3. **Every cited page becomes an opportunity.** One record per page across every answer and every engine, with a 0 to 100 priority, the kind of page it is and the tracked questions whose answers cite it. 4. **The AI review sorts, you decide.** After each data update, the AI review sorts each page by type, scores it, suggests how to act on it and recommends an outreach approach. "Why this classification?" on the page shows its reasoning. The page then reads "AI reviewed", and you approve, reject, park or promote it. Every new project also starts with an auto-promotion rule on: Autopilot finds a contact and drafts a pitch for high-priority pages, and each draft waits for your review. An untouched page that no tracked question cites any more is parked automatically. 5. **Outreach runs from your own inbox.** Researched contacts, a sequence you can edit, and a send that happens only after you approve it. Community threads and directories get a draft you post yourself. 6. **The report shows what changed.** Mentions, citations, share of voice and verified wins, measured before and after the work, so you can show a client what changed. Movement after a placement is reported as association, never as proof that the placement caused it. ## What you bring - The questions buyers ask when they look for what you sell. The setup wizard reads your website and drafts them for each service area it finds. You keep what fits and add your own. - Your competitors, so share of voice has something to compare against. The wizard proposes some from your site, and you add the ones you know. - A mailbox you own, if you want to send outreach from the platform: Gmail or Google Workspace, Outlook or Microsoft 365, or any other mailbox with IMAP and SMTP. You can skip it during setup and connect one later: open Settings, then Mailboxes. - Optionally, Google Analytics 4 and Google Search Console, so the report can show traffic beside visibility. ## What you get back | You ask | Where the answer lives | |---|---| | Are we showing up when buyers ask AI? | Reports: the Overview, Topics and Competitors screens | | Which pages do the answers cite? | Opportunities, sorted by priority | | What should we do today? | Autopilot: Today's deck | | What changed after the work? | Reports: Outcomes and the Impact timeline | | What should we publish on our own site? | Content, then Articles: [The content plan](https://www.sourcesignal.ai/docs/content/the-content-plan.md) | | Can we ask about a project in plain words, or let our own agent in? | [Spot](https://www.sourcesignal.ai/docs/spot-and-agents/spot-the-in-app-agent.md), "Ask Spot" on any page; [your own MCP client](https://www.sourcesignal.ai/docs/spot-and-agents/connect-your-own-agent-over-mcp.md), with a token from Settings, then Agents | Reading this with an agent? [The docs for agents](https://www.sourcesignal.ai/docs/spot-and-agents/the-docs-for-agents.md) says where the markdown copies of every article live. > **Tip:** Metrics across the platform carry a short explainer. Hover over, or tap, the small question mark beside a metric, or a number with a dotted underline, to read what it means. ## Related - [Create your account](https://www.sourcesignal.ai/docs/getting-started/create-your-account.md): Sign up with your email or a Google account, confirm your address, name your workspace and you are in. This article also covers signing in, resetting a password, the account menu and the theme toggle. - [Set up your first project](https://www.sourcesignal.ai/docs/getting-started/set-up-your-first-project.md): The six-step wizard turns one brand into a running project: your brand, research, competitors, prompts, tracking and mailbox. You can leave part way through, and the setup companion brings you back. - [Your first week](https://www.sourcesignal.ai/docs/getting-started/your-first-week.md): What happens in the seven days after setup: the first answers, the first data update, the AI review, your first deck, what to look at each day and each week, and when the report is worth reading. - [What an opportunity is](https://www.sourcesignal.ai/docs/opportunities/what-an-opportunity-is.md): An opportunity is a page an AI engine cited while answering one of your tracked prompts. Each page gets one record that carries its type, its 0 to 100 priority, when and where it was cited, and whether it already names you. --- # Create your account > Sign up with your email or a Google account, confirm your address, name your workspace and you are in. This article also covers signing in, resetting a password, the account menu and the theme toggle. sourcesignal documentation, Getting started. Updated 2026-10-17. The page for people: https://www.sourcesignal.ai/docs/getting-started/create-your-account. Every article: https://www.sourcesignal.ai/llms.txt Your sourcesignal account is you: a sign-in, your name and the workspaces you belong to. You sign up, confirm your email, name your workspace and land in setup for your first project. You come back to the sign-in screen whenever a session ends, and to the account menu whenever you need settings, support or sign-out. Sign-up is open. Anyone with a work or personal email address can create an account, and the first workspace you create starts on a trial (see [Trial and billing](https://www.sourcesignal.ai/docs/account-and-billing/trial-and-billing.md)). ## Before you start - Have an inbox you can read right now. The confirmation link goes there. Throwaway inboxes are refused ("Use a work or personal email address."); free providers such as Gmail are welcome. - Joining a team someone else already set up? Open the invite link they sent you instead of signing up on your own. [Members and roles](https://www.sourcesignal.ai/docs/account-and-billing/members-and-roles.md) covers invites. - Decide how you want to sign in: with a password, or with a Google account (Microsoft and GitHub buttons appear where they are offered). ## Sign up with email and password 1. Open [the sign-up page](https://www.sourcesignal.ai/signup). The heading reads "Create your account"; "Already have one? Sign in" takes you back if you already registered. 2. Fill in First name, Last name (optional), Work email, Company (optional) and a Password of at least 8 characters. 3. Press "Create account". The footnote under the button says it: by creating an account you agree to the [Terms](https://www.sourcesignal.ai/terms) and the [Privacy Policy](https://www.sourcesignal.ai/privacy). 4. The form is replaced by a panel: "Check *your address* for a confirmation link. It signs you in and takes you straight to setup." 5. Open the link in the email. It signs you in and opens Name your workspace. The name and company you typed are already on your profile. Nothing after a minute? Check spam, then press "send the link again" on the panel. You see "A new link is on its way to *your address*." One link goes out per minute per address; asked sooner, the panel says "A link went out in the last minute. Check your inbox and spam." Typed the wrong address? "Change it" takes you back to the form. > **Tip:** Leave the sign-up tab open. If you confirm the link in another tab of the same browser, the waiting tab moves on by itself. ## Sign up with a sign-in provider 1. On the sign-up page, press "Sign up with Google" (or Microsoft or GitHub where offered). 2. Pick the account on the provider's screen. Only your identity is requested: name, email and profile. Nothing else is asked for or stored. 3. You land on Name your workspace. No confirmation email is needed. Cancel on the Google or Microsoft screen and you are back on the sign-in page with no error. If the hand-off fails, the sign-in page says the sign-in didn't complete and asks you to try again or use your email and password. ## Name your workspace A workspace is your organization on sourcesignal. It holds one plan and one bill, your projects (one for each brand you track) and your team. Settings calls it the organization and the account switcher calls it an account. They are the same thing. 1. Under "Welcome aboard" and "Name your workspace.", type a name of 2 to 80 characters. Most people use their company or agency name; the field suggests "e.g. Brightside Agency". 2. Press "Create workspace". 3. You continue into setup for your first project. [Set up your first project](https://www.sourcesignal.ai/docs/getting-started/set-up-your-first-project.md) picks up from here. Three cards under the field say what lives in a workspace: Plan and billing ("One plan and one bill per workspace."), Projects ("One for each brand you track.") and Your team ("Invite teammates right after this."). The footer reminds you: "Joining someone else's team? Open the invite link they sent you instead." Mid-setup, the account menu at the top right shows only your email and "Sign out". "Contact support" joins it once a workspace exists. You can own up to 5 workspaces. "New organization" in the account menu opens the same screen with the heading "Name your new workspace."; a new workspace has its own plan and is billed on its own. ## Sign in Open [the sign-in page](https://www.sourcesignal.ai/login). Under "Welcome back" and "Sign in to sourcesignal." you see: - Provider buttons where offered: "Continue with Google", "Continue with Microsoft", "Continue with GitHub", then "or continue with email". - Fields Email and Password, and the "Sign in" button. - "Forgot your password?" under the form, and "New here? Create an account" above it. A "Last used" badge marks whichever method this browser tried last, so you do not have to remember which one your account uses. It is a per-browser hint, nothing more. Already signed in? You are sent straight to the page you asked for. If your email is not yet confirmed and the password is right, sign-in sends a fresh confirmation link instead and tells you so: "Your email is not confirmed yet. A new confirmation link is on its way to *your address*." > **Heads up:** A confirmation link may not finish signing you in when it opens in a different browser from the one you signed up in, or when a mail scanner opened it first. The sign-in page then says "That link did not sign you in on this device." Your address is usually confirmed by then, so sign in with your password. If the link had expired, signing in sends a new one. ## Reset a password 1. Press "Forgot your password?" on the sign-in page. The screen reads "Reset your password" with "We'll email you a link that signs you in and lets you set a new one." 2. Enter your email and press "Email me a reset link". You see "If that address has an account, a reset link is on its way. It expires after an hour." ("Remembered it? Sign in" takes you back.) 3. Open the link within the hour. It signs you in and opens "Choose a new password" for your address. 4. Enter a New password of at least 8 characters and press "Set password & continue". You land in the app. An expired or already used link sends you to the sign-in page with "That reset link has expired or was already used. Use Forgot your password to get a new one." Ask too often and the form says "Too many attempts. Try again in a minute." Signed in and want to change your password or email, or end sessions on other devices? That lives in Settings, then Profile: see [Profile and sign-in](https://www.sourcesignal.ai/docs/account-and-billing/profile-and-sign-in.md). ## The account menu On every app page, the circle at the top right carries the first letter of your email. Open it to see: - Your email, and one line on your plan where there is one: the days left on your trial, "Cancels on" a date or "Full access until" a date. - A short usage preview (your plan, your team seats and, when your plan caps them, tracked prompts) with a link to Billing. - "Account settings", which opens Settings on the Organization tab. - "Support", which opens the support drawer. A small dot on the circle and a "new" count beside the item mean a reply is waiting. - "New organization", which starts another workspace. - "Sign out". Signing out also clears any half-typed support message in this browser, so the next person on a shared machine never finds it. If you belong to more than one workspace, a pill naming the active one sits at the top left, by the sourcesignal logo. [Organization settings](https://www.sourcesignal.ai/docs/account-and-billing/organization-settings.md) covers switching and starting another workspace. ## Theme Next to the account menu, a moon or sun icon switches between the light theme (the default) and dark. Its label reads "Switch to dark theme" or "Switch to light theme". The choice is remembered in this browser and applied before the page paints, so there is no flash on your next visit. It is per browser, not per account. ## Good to know - A password set at sign-up or on Choose a new password needs at least 8 characters. Changing or setting one later from Settings asks for at least 10. - Signed in with Google? You can add a password as a second way in from Settings, then Profile. Single sign-on keeps working. - Your first name and company from sign-up are on your profile from the start. Teammates see your name beside your work; change it any time under Settings, then Profile. - The confirmation link and the reset link both sign you in. When one works, you go straight on without signing in again. ## Related - [Set up your first project](https://www.sourcesignal.ai/docs/getting-started/set-up-your-first-project.md): The six-step wizard turns one brand into a running project: your brand, research, competitors, prompts, tracking and mailbox. You can leave part way through, and the setup companion brings you back. - [What sourcesignal does](https://www.sourcesignal.ai/docs/getting-started/what-sourcesignal-does.md): sourcesignal tracks the questions buyers ask AI engines, shows you the pages those answers cite, and helps you work your brand onto them. - [Profile and sign-in](https://www.sourcesignal.ai/docs/account-and-billing/profile-and-sign-in.md): Your profile holds the settings that belong to you, not to a workspace: your name, photo and timezone, the email and password you sign in with, and the devices signed in as you. - [Members and roles](https://www.sourcesignal.ai/docs/account-and-billing/members-and-roles.md): Owners, admins and members can each do different things in a workspace. This page covers inviting a teammate or another user, choosing their role and projects, sharing ownership, leaving and how seats count. - [Trial and billing](https://www.sourcesignal.ai/docs/account-and-billing/trial-and-billing.md): Your first workspace starts with a 14-day free trial, and you need no card. This page covers what happens when the trial ends, and how to pay, change plan or cancel your subscription from Settings, then Billing. --- # Set up your first project > The six-step wizard turns one brand into a running project: your brand, research, competitors, prompts, tracking and mailbox. You can leave part way through, and the setup companion brings you back. sourcesignal documentation, Getting started. Updated 2026-10-22. The page for people: https://www.sourcesignal.ai/docs/getting-started/set-up-your-first-project. Every article: https://www.sourcesignal.ai/llms.txt The setup wizard turns one brand into a running project. You name the brand and its website, the autopilot reads the site and proposes service areas, competitors and the buyer questions to track, you keep what fits, tracking switches on, and you connect the mailbox outreach sends from. One decision per screen, six screens. You meet it right after you create your workspace, and again whenever you press "New project". The wizard remembers where you left off, so you can leave part way through and come back later. ## Before you start A workspace is your organization on sourcesignal. It holds one plan, your projects (one for each brand you track) and your team. If you are signed in with no workspace yet, the first screen is "Name your workspace." Type a name, press "Create workspace", and step 1 opens. See [Create your account](https://www.sourcesignal.ai/docs/getting-started/create-your-account.md). The rail on the left lists the six steps in order: "Your brand", "Research", "Competitors", "Prompts", "Tracking", "Mailbox". Each finished step keeps a check mark and one fact, such as the brand name, "N areas" or "Live". Under the steps, the plan card reads "You're on ", shows the trial clock while a trial runs, and shows the one limit the current step spends. "Compare plans" opens every plan side by side. On a phone the rail becomes a step bar with an "All steps" disclosure. ## Step 1: Your brand On your first project the screen is titled "Let's track your first brand."; on a later one, "Add a brand to track". You fill in: - "Brand name" and "Website", both required. The website is read on the next step. - "Target markets", in priority order, from United States, United Kingdom, Ireland, Australia, Canada, New Zealand and Worldwide. The first country is where every prompt is checked; the rest shape the plan. With the United States on the list you can add the states you serve. At least one market is required. See [Markets and locations](https://www.sourcesignal.ai/docs/tracking/markets-and-locations.md). - "Project name and industry", a disclosure. The project name defaults to the brand name. "Industry or niche" starts on "Auto-detect"; it is context for AI summaries and briefings and does not change what is checked. Press "Create project". The wizard moves to Research by itself. ## Step 2: Research The site read starts on arrival, no click needed. The title reads "Reading " over a log of real events: your home page being read, "Mapping service areas…", "Found N service areas", "Industry detected: ". At rest the title becomes "Your site maps to N service areas". Each area becomes one group of buyer questions. - Hover an area chip to read why the AI proposed it. Remove the ones you do not offer. - Type your own in "Add an area you serve…". A plan holds up to 12 areas. - Change the detected industry with "Change" and "Save". - "Re-read the site" reads again and keeps the areas you typed. Press "Draft the plan from N areas". The plan keeps drafting while you move on to Competitors. Once a plan exists, the button reads "Continue", with "Draft again (replaces the competitors and prompts)" beside it. A quiet "Deep dive" card sits under the read. It is an optional deeper read of your site, page by page and sitemap first, up to your plan's page budget. It drafts prompts for each service area your pages cover and proposes competitors. You approve its proposal before anything is added. On a trial the card reads "Choose a plan to unlock"; on a plan without it, "Upgrade to "; on a plan that carries it, "Run the deep dive". See [Plans and limits](https://www.sourcesignal.ai/docs/account-and-billing/plans-and-limits.md). ## Step 3: Competitors "Who buyers compare you with" lists your own brand pinned at the top, then each proposed competitor with a checkbox, its website and an info hint, "Why ". - Uncheck a competitor to drop it. Your unchecks are saved with the draft. - "Add a competitor" by name and website. It is written to the project at once, tagged "Added by you", with a "Remove" button. - If the proposal failed, add the ones you know or continue. More are added as they show up in AI answers. Press "Continue with N competitors" or "Continue without competitors". See [Competitors](https://www.sourcesignal.ai/docs/tracking/competitors.md). ## Step 4: Prompts "What buyers ask AI" shows one accordion per group. The first group is open, the rest are folded. Each group has a checkbox for all its questions; each question has its own checkbox and a buying-stage chip (Awareness, Problem-aware, Research, Comparison, Consideration, Purchase). - Check or uncheck questions and whole groups. A group with nothing checked is left out. - Type your own question in "Add a question…" inside a group. It carries no stage chip until the question has been tracked. - Add a group of your own research in "A group of your own…" and press "Group". A plan holds up to 12 groups. - Read the count line, "N prompts · M competitors". The plan card counts your checks live against your plan's prompt allowance, which is one pool across all your projects. Press "Add and start tracking". It writes the checked prompts and kept competitors and switches tracking on in the same press. One press adds up to 100 prompts and 15 competitors; add the rest afterward on the Prompts and Competitors pages. The button waits while the plan is still drafting, when nothing is checked, or when your selection is over the pool. See [Prompts](https://www.sourcesignal.ai/docs/tracking/prompts.md). ## Step 5: Tracking Tracking asks the AI engines your questions on a schedule and reads back what they answer and cite. Switching it on takes a minute or two. The screen logs the real stages, from "Preparing your tracking feed…" to "Tracking feed live", then reads "Tracking is live" with "N prompts are being checked." The next step opens by itself. The engines are chosen per project. A new project starts on ChatGPT, AI Overviews and Perplexity; change the set any time in the project's settings, under AI engines. See [AI engines](https://www.sourcesignal.ai/docs/tracking/ai-engines.md). If tracking does not switch on, the title reads "Tracking didn't switch on yet" with the reason. Your plan is saved either way. Press "Try again", or "Continue to the mailbox" and switch tracking on from the end screen later. If your workspace has no tracking feed yet, "Contact support" opens the support drawer prefilled with "Switch tracking on for ". ## Step 6: Mailbox "Connect the mailbox that sends your outreach" offers "Connect Gmail" (Gmail or Google Workspace), "Connect Outlook" (Outlook or Microsoft 365) and "Connect other email" (any other provider, with IMAP and SMTP: you type the mailbox's details on the connection screen). You go to the connection screen and come straight back to this step, where "Confirming your mailbox" waits for the provider, usually a few seconds. After two minutes it offers "Check again". > **Note:** Connecting a mailbox is not how you sign in. It gives sourcesignal permission to send from that inbox, and your Google or Microsoft login stays as it is. Nothing sends until you launch it, or Spot or your agent does where you set "Emails and posts to publishers" to Apply at once. Disconnect it any time in Settings, Mailboxes. When the address has no sender name or signature yet, owners and admins then see "Sign your pitches": the sender name and signature every pitch from this address ends with. Fill them in and press "Save and finish", or "Skip for now". Once connected, the step reads "Your mailbox is connected" with "Connect a different mailbox" beside it, and "Finish" moves you on. You can skip the whole step with "Skip for now". Outreach waits for a mailbox, and the console's mailbox card and the setup companion keep the ask. The skip is remembered for this project in this browser. On a later project in a workspace that already has mailboxes, owners and admins pick the address instead: "Which address sends for ?", then "Send from " or "Connect another mailbox". A member reads which address an admin picked and presses "Finish". See [Mailboxes](https://www.sourcesignal.ai/docs/account-and-billing/mailboxes.md). ## The end screen "You set up ." says what you set up and what happens next: - "Now": your prompts are queued with the project's engines. - "Within hours": the first answers usually land within minutes to a few hours. The platform checks hourly for the first two days, then every morning at 06:00 UTC. - "After that": pages the answers cite get a 0 to 100 priority. Every new project starts with a default rule on: Autopilot finds a contact and drafts a pitch for each page the rule picks up, the highest priority first, up to 30 a day. You review each one, and nothing sends until you launch it, or Spot or your agent does where you set "Emails and posts to publishers" to Apply at once. Without a mailbox, the line says Autopilot drafts pitches once outreach has one. Anything still left has one line and one action: "Switch tracking on", "Check on tracking", "Connect a mailbox", "Pick the address" or "Sign your pitches". A tally reads "What you set up: N areas · N competitors · N prompts · N mailbox". Press "Open the console". What to do once answers land is in [Your first week](https://www.sourcesignal.ai/docs/getting-started/your-first-week.md). ## Finish later and coming back "Finish later" sits in the top bar from Research on and takes you to the console. It steps aside while tracking is switching on. Three ways lead back: the setup companion's "Back to the setup wizard", "Continue setup" on the project's card in the project list, and "Continue setting up " above the form on step 1 of a new setup. The wizard reopens at the step your project is actually at: a running read is picked up, a saved draft restores your unchecks, a project with prompts but no feed opens on Tracking. Earlier steps are one click away in the rail, or Back, as long as nothing is running. They wait, with the reason on hover, while the site is being read, the plan is drafting or tracking is switching on. > **Tip:** Changing the brand domain on "Your brand" ("Edit", then "Save domain") starts the plan over from the new site. Once tracking is live the domain is locked to your tracking feed and changes only in project settings. ## The setup companion On every page of a project that is not fully set up, a panel floats bottom right (not on phones). It is titled "Set up " and lists the live steps as links: "Your brand", "Add prompts to track", "Add competitors", "Activate AI answer tracking", "First answers land" and "Connect a mailbox · optional". Done steps are checked, including steps finished elsewhere or by a teammate, which the panel picks up when you come back to the window. Its button reads "Back to the setup wizard" until the project is running, then "Open the full setup". "Minimize" folds it to a pill, "Setup 1/3", counting the three core steps: prompts added, tracking active, first answers landed. Competitors and a mailbox never block it. Once the core is done you can remove the panel with "Remove (setup is complete)". ## The full setup checklist "Get set up" is the same setup on one page, for whoever prefers to add prompts by hand or wants to see where tracking got to. Open it from the companion's "Open the full setup". - "Your brand": the brand card with "Edit" and "Brand domain". - "Add the prompts you want to track": "Add prompt" one at a time, or "Bulk import" to paste a list, one per line. Each row has "Remove". - "Review your competitor set": the roster with an add form and "Remove". - "Activate AI answer tracking": "Set up my tracking", then the same stages as the wizard until "Tracking feed live". - "Wait for the first answers": "Check now" asks the tracking feed right now. The result reads "Nothing has arrived yet" or "N answers landed. The console is waking up." - "Connect a mailbox", marked "Optional". The footer says it plainly: AI insights, briefings and the autopilot all need answers, and they wake up as soon as the first ones land. ## Good to know - Everything stays editable afterward. Prompts and competitors have their own pages, markets and engines live in the project's settings, and mailboxes in Settings, Mailboxes. - Your plan's prompt allowance is one pool across all projects; a prompt counts once however many engines answer it. - A buying stage is a first read. Once the prompt is tracked it is judged again, so the chip can change. - Removing a competitor keeps its history for reporting. It may be proposed again if it keeps appearing in your AI answers. - A member assigned to specific projects cannot create a project. An owner or admin adds it and gives access. - The tracking feed is an Ahrefs Brand Radar report the platform creates for you. It is named in project settings, not in the wizard. See [The answer feed and data updates](https://www.sourcesignal.ai/docs/tracking/the-answer-feed-and-data-updates.md). ## Related - [Create your account](https://www.sourcesignal.ai/docs/getting-started/create-your-account.md): Sign up with your email or a Google account, confirm your address, name your workspace and you are in. This article also covers signing in, resetting a password, the account menu and the theme toggle. - [Your first week](https://www.sourcesignal.ai/docs/getting-started/your-first-week.md): What happens in the seven days after setup: the first answers, the first data update, the AI review, your first deck, what to look at each day and each week, and when the report is worth reading. - [Prompts](https://www.sourcesignal.ai/docs/tracking/prompts.md): The Prompts page holds the questions buyers ask AI engines and shows, for each one, how often the answers named you over the last 30 days. Add, import, classify, remove and restore them here. - [Markets and locations](https://www.sourcesignal.ai/docs/tracking/markets-and-locations.md): Choose the countries you sell into, in priority order. The first country is where every prompt is checked; the rest, and any US states you serve, shape how your tracking plan is phrased. - [Mailboxes](https://www.sourcesignal.ai/docs/account-and-billing/mailboxes.md): Outreach sends from a Gmail, Outlook or any other mailbox with IMAP and SMTP that your organization connects. This page covers connecting one, how its emails sign off, its sending limits, and what to do when it reads Reauth required, stops sending or you retire it. --- # Your first week > What happens in the seven days after setup: the first answers, the first data update, the AI review, your first deck, what to look at each day and each week, and when the report is worth reading. sourcesignal documentation, Getting started. Updated 2026-10-22. The page for people: https://www.sourcesignal.ai/docs/getting-started/your-first-week. Every article: https://www.sourcesignal.ai/llms.txt Your project is running. The wizard's end screen told you what comes next: the first answers usually land within minutes to a few hours, the deck fills in, and nothing sends until you launch it, or Spot or your agent does where you set "Emails and posts to publishers" to Apply at once. This article walks through the seven days after that screen: what arrives when, what you see on the console, and what is worth your time each day and each week. Autopilot checks the tracking feed on its own, reviews what arrives and deals you the decisions that are yours. Your part is a few minutes a day on the deck, and a look at the report once it has something to say. ## The first hours Press "Open the console". While no answer has landed, the masthead reads "Tracking is live. First answers on the way." and the deck leads with a card headed "First run in progress": "Your questions are queued with the engines". It lists what is already done and names the next automatic check. For the first two days the feed is checked every hour, at :15 UTC. Once the first answers are in, or two days after tracking starts, the daily data update at 06:00 UTC takes over. You can also press "Check now" on the card to ask the feed straight away. If nothing has arrived the result says so, and a short cooldown applies before the next press. When the first answers land, the owners and admins get one email about it, once per project and usually within two hours: it names the first page worth a decision once one is scored, says who AI named, and its button opens your deck. The floating setup checklist reads "Tracking is live. Waiting for the first answers." until its step "First answers land" is checked. Then it closes on its own; if you had minimized it, opening it reads "Answers are flowing. You can remove this." > **Note:** A quiet first morning is normal. Each AI engine is asked on its own cadence. Once the first check has read it back, the console's Schedule card says what the feed runs, for example "ChatGPT, Gemini weekly · Perplexity daily". You choose which engines a project tracks, not how often each is asked. ## The first data update Every update is recorded on the Imports tab of the Data step, under Settings: one row per run with its day, "Scheduled" or "Manual", and how many of the three datasets arrived (cited pages, visibility and AI answers). The note beside "Run import now" shows the last update and a countdown to the next. Three outcomes matter in the first week: - "Ready": the run finished and saved its data. - "Nothing new": the feed had nothing newer than the answers already held. A quiet morning, not a failure. - "Failed": the row says in plain words what is missing, and a notice at the bottom of the screen offers "Retry" (or "Open settings") and "Contact support". It clears on its own once a run reaches the feed again. [The answer feed and data updates](https://www.sourcesignal.ai/docs/tracking/the-answer-feed-and-data-updates.md) covers every state. ## The AI review Whatever an update brings in is then reviewed by AI without you touching anything. If the review leaves pages unfinished, a strip above the table on the Opportunities page says how many are waiting ("N pages are waiting for AI review") with a "Run the AI review" button; running it again is safe, nothing is done twice. A page moves from "New" to "AI reviewed" once the review has been through it. Each cited page gets a 0 to 100 priority for outreach. "Why this score" on its deck card names the reasons for that page. Your prompts change a little too: the buying stage you saw in the wizard is a first read, and it can move once the prompt is tracked. The deck narrates it. While a run is going, the lead card names the step running and when to look again. When the deal lands you read "Autopilot dealt you N decisions", the strongest first. If the first pass found nothing worth a decision yet, the deck says "Your project is set up" and the autopilot keeps checking on its schedule. ## Your first deck Every new project starts with the default auto-promotion rule on. Once pages are scored, Autopilot finds a contact and drafts a pitch for each page the rule picks, the highest priority first, up to 30 a day, and each draft reaches your deck as an "Approve draft" card. Drafting needs a sending mailbox. If you skipped that step, a "Connect a mailbox" card leads the deck: connect Gmail, Outlook or other email from it, or later under Settings, then Mailboxes. Put it away with "Not now" and a header chip keeps the ask, counting the pages waiting. The cards you are likely to meet first: - "Approve draft": the opening email as it will land, with your signature, and its follow-ups. Press A, "Mark reviewed", and it moves to Outreach, Ready to launch. Approving never sends. - "Promote?": a scored page the autopilot left to you, with its priority, how many answers cite it and "Why this card". Press A to promote it; a contact is found and the draft comes back for your approval. - "New competitor": a brand AI answers name alongside you that you do not track yet. Track it or ignore it. The deck deals 10 new pitches a day by default. Click the count in the deck header to open "Today's deck" and set the daily decision budget anywhere from 1 to 30. [The deck](https://www.sourcesignal.ai/docs/autopilot/the-deck.md) describes every card and key. > **Tip:** Nothing you approve on the deck is sent. Open Outreach, then Ready to launch, and launch the sequences yourself. Each follow-up waits its days and only sends if nobody has replied. See [Reviewing and launching](https://www.sourcesignal.ai/docs/outreach/reviewing-and-launching.md). ## What to look at each day A daily visit is short. 1. Read the console's first lines: your good news, then the ask ("N decisions are yours today, about N minutes."). The count scrolls to the deck. 2. Open "Since your last visit" in the deck header. Your first visit covers the last 7 days: responses read, new opportunities, drafts prepared, replies and wins, each chip a link to its rows. 3. Clear the deck. Replies come first and are always dealt; once it is clear, "Pull another stack" deals five more pitches. 4. Launch what you approved under Outreach, Ready to launch. 5. Glance at the rail's foot. It says what Autopilot is doing right now, or that all is quiet; "Engine room" opens the run in progress and the steps your data moves through, each with a status. "What the autopilot did" on the console logs every import, draft, send and reply, day by day. ## What to look at each week - The "Your week" card in the rail: AI share of voice with its change in points, replies and emails sent this week, wins this month. Before enough history it reads "Your share-of-voice trend appears after a few days of imports." - The weekly briefing, written every Monday at 07:00 UTC: a read of your last 7, 30 and 90 days that says what moved, what landed and what is worth a look, kept with your reports under Internal analysis. - The Schedule tab of the Data step: everything that runs for the project, when it runs next and when it last ran. - The Prompts and Competitors pages: prune questions that do not fit, and keep or drop the brands the autopilot suggests. ## When the first report is worth reading Reports compute every figure over complete days, so a window always ends yesterday and today's checks are shown separately as "data collection in progress". Open Reports after your first few updates and read the Overview's cover sentence: how often your brand was named in the tracked AI answers over the period. A missing number is a dash with its reason, never a zero. What to expect in week one: - A topic with too few checks reads "Insufficient data", not a percentage. Topics fill in as checks accumulate. - A comparison needs the earlier period on record, so "Vs previous period" stays empty until enough history has built up; on "All time" the picker says "Nothing to compare". The report opens on "Last 90 days". - The Impact timeline needs a verified win, and a win needs about five weeks before its before-and-after can be read. Newer wins read "Early data". - Traffic & search shows "Not connected" until you connect Google Analytics 4 or Search Console in the project's settings, under Google traffic. The report is worth sharing once the Overview has a period you can compare and the topics you care about carry numbers rather than "Insufficient data". Until then, the weekly briefing is the better read, from its first Monday on. [Reports overview](https://www.sourcesignal.ai/docs/reports/reports-overview.md) explains each screen. ## Related - [The deck](https://www.sourcesignal.ai/docs/autopilot/the-deck.md): Your deck deals the decisions only you can make, one card at a time. This article covers every card, the keys, undo, the daily budget and what a cleared deck means. - [The answer feed and data updates](https://www.sourcesignal.ai/docs/tracking/the-answer-feed-and-data-updates.md): Connect one Ahrefs Brand Radar report, managed for you or your own, and the daily data update brings in AI answers, cited pages and visibility every morning. The Data tab keeps every run, the full schedule and the way to fix a feed that stops. - [Set up your first project](https://www.sourcesignal.ai/docs/getting-started/set-up-your-first-project.md): The six-step wizard turns one brand into a running project: your brand, research, competitors, prompts, tracking and mailbox. You can leave part way through, and the setup companion brings you back. - [Reports overview](https://www.sourcesignal.ai/docs/reports/reports-overview.md): Reports shows how visible your brand was in AI answers over a period and what your work delivered. It is one live report with six screens and one filter bar, and every window ends at the last complete day. - [Reviewing and launching](https://www.sourcesignal.ai/docs/outreach/reviewing-and-launching.md): Every drafted pitch waits for you twice: once under Needs review to read and sign off, once under Ready to launch to send. Nothing goes out until you launch it, or Spot or your agent does where you set "Emails and posts to publishers" to Apply at once. --- # Prompts > The Prompts page holds the questions buyers ask AI engines and shows, for each one, how often the answers named you over the last 30 days. Add, import, classify, remove and restore them here. sourcesignal documentation, Tracking. Updated 2026-10-18. The page for people: https://www.sourcesignal.ai/docs/tracking/prompts. Every article: https://www.sourcesignal.ai/llms.txt Prompts is where you decide what visibility means for a project. Each prompt is a question your buyers ask AI assistants. Once you add one, every answer to it is captured from then on, and its row tells you how often AI names you, where it places you, which competitor it names most, which engines answer it and which pages the answers cite. You come here to add or import questions, to read where you stand per question and per topic, to find the gaps worth working on, and to keep the list tidy. ## What you see The lead band, "Your mention rate across prompts · last 30 days", carries one large percentage and the line "Named in N of M answers to K prompts". Three tiles sit beside it: - **Never named**: answered prompts that never named you in the window. Each is a question where AI recommends someone else, or no one. - **Biggest gap**: the one prompt most worth working on. "Biggest gaps first" under Sort lists them all. - **Strongest**: the prompt whose answers name you most often, as a share of its answers. The filter bar holds "Search prompts" (it reads the question, topic, cluster and tags), "Topic", "Buying stage", "Organize" (Group by topic, Group by buying stage, No grouping) and "Sort" (Highest intent, Your mention rate, Biggest gaps first, Most answers, Most opportunities, Newest first, A–Z). Prompts are drawn in group cards, one per topic or buying stage. Each header shows the group name, "N prompts", "avg intent N" and "you're named in X% of N answers". Topic groups are ordered by their highest-intent prompt, with "Topic pending" last; stage groups run from Purchase down to Unclassified. On wide screens, a sidebar titled "Topics · N" or "Buying stages · N" lists every group with its prompt count and your mention rate as a bar; click one to jump to it. On smaller screens, a row of chips above the list does the same. ### A prompt row Each row has a checkbox, the question, its chips (the topic when the list is not grouped by topic, or "Topic pending"; the buying stage; "Intent N" or "Unclassified"; and "N keywords" when search terms stand behind it) and four cells. | Cell | What it shows | |---|---| | You | "X% of N" answers over the last 30 days, and where you usually sit: "usually the lead pick", "usually one of a list" or "usually a footnote". | | Top competitor | The competitor named most in the window, "in N answers", or "None named". | | Engines | A stack of engine logos, with "+N" when more than five answer it, or "Not answered yet". Open the stack for each engine's answer count and newest answer. | | Cited pages | "N pages →", which opens exactly those pages in Opportunities, or "0". | "Details · N answers" opens the rest: your cluster and tags, what the AI review observed ("Brand-mention", "Source-only", "N competitors"), "Why this stage" with its one-line reason, a per-engine summary such as "Last captured today · ChatGPT X/N" (each engine's answers that named you out of its total), and the captured answers, newest first, each with "View answer →". > **Note:** A dash in the You column means no answer was stored for that prompt in the last 30 days. It is not a zero. "0%" means answers were stored and none named you. An answer that only cites your site does not count as naming you. ### Topics, clusters and tags A topic is the group the AI puts a prompt in. Grouping runs when prompts are added and after every import; until then the row reads "Topic pending". A cluster and tags are your own labels: one optional cluster and any comma-separated tags, shown under Details and read by the search box. ### Buying stage and intent The buying stage says where a question sits in the journey: Awareness, Problem-aware, Research, Comparison, Consideration, Purchase or Unclassified. It can be revised once the prompt has been tracked for a while, so a first read is not final. Intent, shown as "Intent N" on a 0 to 100 scale, estimates how much being named in this question's answer would be worth. It scores the question, not your chances of being named. "Brand-mention" means AI names brands in these answers, so outreach to the cited pages can work. "Source-only" means AI cites sources but names no brand: aim to become the cited source rather than pitching. ## How to add prompts 1. Press "Add prompt". The dialog "Add a prompt" takes the question, an optional cluster and comma-separated tags. 2. To paste a list, press "Bulk import": one prompt per line, with ` | cluster | tag1,tag2` appended to set labels. Duplicates are skipped. Press "Import". Rows that fail stay in the box with their reason; everything else goes in. 3. With an Ahrefs Brand Radar report connected in the project's settings, press "Import from Ahrefs" to pull the report's prompt list now. The toast says what was imported, restored or removed, or "Already in sync." 4. Wait for the next data update. Then the row fills in and "Details" opens the captured answers. See [The answer feed and data updates](https://www.sourcesignal.ai/docs/tracking/the-answer-feed-and-data-updates.md). To update a prompt's stage and intent now, press the row's Classify control (the sparkle icon on wide screens, "Classify" or "Re-classify" on smaller ones), or check several and press "Classify selected". Topics are grouped separately, automatically. "Edit" (the pencil icon) opens the question, cluster and tags inline with "Save changes" and "Cancel". ## How to find your biggest gaps 1. Set "Organize" to "Group by buying stage" or "Group by topic", and "Sort" to "Biggest gaps first". 2. Read the lead band's "Never named" and "Biggest gap", then each group header's "you're named in X% of N answers". 3. Click "N pages →" on a prompt to open the pages AI cited for it in Opportunities. Every one is an [opportunity](https://www.sourcesignal.ai/docs/opportunities/what-an-opportunity-is.md) you can work. The "N keywords" chip opens "Keywords behind this prompt", and "Manage keywords →" lands on [Keywords](https://www.sourcesignal.ai/docs/tracking/keywords.md) filtered to it. ## How to remove and restore a prompt 1. Press the row's "Delete" control (the trash icon) and confirm "Remove this prompt?" with "Remove". The dialog says what happens: its opportunity links are removed, affected opportunities are re-scored, and imports will not re-add it. 2. The prompt moves to "Removed prompts (N)" at the foot of the page, struck through and marked "Removed here" with the date. 3. Open that section and press "Restore" to bring it back. A restore counts against your prompt allowance like an add. Prompts the Brand Radar report no longer lists land in the same section as "Removed from the Brand Radar report". You restore them the same way. ### Prompts paused by your plan After a move to a smaller plan, prompts the plan no longer covers are paused, not deleted. They sit under "Paused by your plan (N)" with the line "Kept with everything they found. An owner or admin brings them back from Billing, under What stays." There is no Restore button here: the way back is that picker, or a plan with room. See [Plans and limits](https://www.sourcesignal.ai/docs/account-and-billing/plans-and-limits.md). ## Re-reading your site for suggestions "Re-read site" reads your site again and compares it with what you track. The menu shows the last read ("Last read N days ago · N pages", or "Not read since setup") and offers two reads: - **Re-read homepage** reads the homepage again and shows what changed against your plan. Every plan has it. - **Deep dive** reads up to your plan's page count, sitemap first, drafts prompts for each service area your pages cover and proposes competitors. It comes with Growth, Agency and Enterprise; on Brand and during a trial it is shown with a lock. The result lists the service areas ("N not in your plan"), the proposed prompts by area with stage and intent chips and a "New" or "Already tracked" tag, and proposed competitors tagged "New" or "On your roster". Uncheck what you do not want and press "Add selected", or "Not now". Nothing is added until you press it. Competitors added here join the roster described in [Competitors](https://www.sourcesignal.ai/docs/tracking/competitors.md). ## Limits | What | Limit | |---|---| | Tracked prompts, pooled across every project in the workspace | 30 on Brand, 200 on Growth, 750 on Agency; custom on Enterprise | | Prompt text | Up to 500 characters; a cluster up to 60; each tag up to 40 | | "Classify selected" | Up to 1,000 checked prompts per press | | One "Add selected" from a site read | Up to 100 prompts and 15 competitors | | Deep dive pages | 300 on Growth, 1,000 on Agency; Brand has the homepage re-read only | | Mention rate, position and top competitor | The last 30 days of stored answers, the newest 5,000 at most; cited pages count all time | > **Heads up:** Past the pooled allowance, adding, importing or restoring is refused with a line such as "Your Growth plan tracks 200 prompts across all projects (200 in use). Upgrade or remove some to add another." When an Ahrefs import hits it, the rest of the report waits. ## Good to know - A prompt is checked in the first country of your target markets. See [Markets and locations](https://www.sourcesignal.ai/docs/tracking/markets-and-locations.md). - Which engines answer depends on the engines switched on for the project. See [AI engines](https://www.sourcesignal.ai/docs/tracking/ai-engines.md). - The "Cited pages" count leaves out pages the Opportunities table hides by default (parked, rejected, already live), so the number and the list agree. - You can ask Spot to find prompts by words, topic, stage, and whether you or a competitor were named. See [Spot, the in-app agent](https://www.sourcesignal.ai/docs/spot-and-agents/spot-the-in-app-agent.md). ## Related - [Keywords](https://www.sourcesignal.ai/docs/tracking/keywords.md): Track the search terms your buyers type beside the questions they ask AI, per market, and link each one to the prompts it stands behind. - [Competitors](https://www.sourcesignal.ai/docs/tracking/competitors.md): Build the roster of brands AI names instead of you, compare your share of AI answers with theirs, and see where each one wins and which sources carry it. - [The answer feed and data updates](https://www.sourcesignal.ai/docs/tracking/the-answer-feed-and-data-updates.md): Connect one Ahrefs Brand Radar report, managed for you or your own, and the daily data update brings in AI answers, cited pages and visibility every morning. The Data tab keeps every run, the full schedule and the way to fix a feed that stops. - [What an opportunity is](https://www.sourcesignal.ai/docs/opportunities/what-an-opportunity-is.md): An opportunity is a page an AI engine cited while answering one of your tracked prompts. Each page gets one record that carries its type, its 0 to 100 priority, when and where it was cited, and whether it already names you. - [Plans and limits](https://www.sourcesignal.ai/docs/account-and-billing/plans-and-limits.md): What Brand, Growth, Agency and Enterprise include, how each limit counts and what happens when you reach one. When a plan is smaller than what you track, nothing is deleted: you choose what stays and the rest is paused. --- # Keywords > Track the search terms your buyers type beside the questions they ask AI, per market, and link each one to the prompts it stands behind. sourcesignal documentation, Tracking. Updated 2026-10-22. The page for people: https://www.sourcesignal.ai/docs/tracking/keywords. Every article: https://www.sourcesignal.ai/llms.txt The Keywords page holds the search terms your buyers type into a search box, tracked per market, beside the prompts they ask AI assistants. A keyword carries demand: how many people search it each month, what a click costs and how crowded the term is. A prompt carries the answers. Keep both, link each keyword to the prompts it stands behind, and you can read demand and answers together. You use it when you set up a project, when you bring in a list from your keyword tool, and whenever you want to know which search terms sit behind a question you track. The page sits in your project's side rail beside Prompts and Competitors. You can also press g then k to open it (see [Navigation and keyboard shortcuts](https://www.sourcesignal.ai/docs/reference/navigation-and-keyboard-shortcuts.md)). ## A keyword beside a prompt - A **prompt** is a question your buyers ask an AI assistant. Every answer to it is captured and read. See [Prompts](https://www.sourcesignal.ai/docs/tracking/prompts.md). - A **keyword** is a term your buyers type into a search box. It is tracked in one market and carries up to three numbers from your keyword tool or your own typing. No AI engine is asked about a keyword. Linking the two says "this is what people search when they mean this question". Nothing on the Keywords page changes a score, an opportunity or a report. ## What you see Under the lead line "The search terms you track beside your prompts, per market." sit three views. ### Keywords The list opens with "Keywords · N tracked" and "N stand behind a prompt; M have a volume.", a "Find a keyword…" box and a sort: "Highest volume", "Closest to buying", "Most prompts", "Newest first" or "A–Z". Each row shows: | Column | What it shows | |---|---| | Keyword | The term. A market chip appears when the keyword has a market of its own; a keyword on "Project default" wears none. | | Stage | The buying stage set when the keyword was added. No chip while it is not set. | | Volume | How many times a month people search the term, as its provider estimates it. A modeled figure for the keyword's market, not measured traffic. | | CPC | What advertisers pay per click, in the provider's currency (usually US dollars). A higher cost usually means buyers rather than browsers. | | Difficulty | How hard it is to rank for the term in web search, 0 to 100, as its provider scores it. Not a measure of AI answers. | | Prompts | The tracked prompts the keyword stands behind. | CPC and Difficulty appear from wide screens; on a phone each keyword is a card with every number. An empty number reads "Not fetched", never 0. Hover a filled volume to see where it came from: "Numbers typed in" or "Numbers from your CSV". After 60 rows, "Show N more · M not shown" draws up to 60 more at a time. The find box matches the term, a market code or the text of a linked prompt. That is how "Manage keywords →" on the Prompts page lands here filtered to the keywords behind one prompt. ### By prompt "By prompt · N with keywords" reads the same links the other way round: one card per tracked prompt, its keywords as pills with the term, the market when the keyword has its own and the volume when there is one. Prompts with the most keywords come first. Two folded groups follow: "Prompts with no keyword yet (N)" and "Keywords behind no prompt (N)". Until you link something it says "No keyword stands behind a prompt yet". ### Archived Every removed keyword, struck through, with its market chip, why it left and when, and a "Restore" button. The reasons shown are "You removed it", "Its source dropped it" and "Paused by the plan". With nothing removed it reads "Nothing archived". ## Keyword stage A keyword's stage is one of Awareness, Problem-aware, Research, Comparison, Consideration or Purchase, or "Not set". You set it when you add the keyword, by hand or from the stage column of your file, and it stays as you set it. The prompts it stands behind carry their own stage, set by the platform and revised as the prompt is tracked, so the two can differ. Sort by "Closest to buying" to read Purchase first. ## How to add keywords The platform can also suggest keywords from the prompts you track: you keep, drop or rewrite each one in a preview, nothing is added until you press Add, and a term you rewrite is added as your own. ### One at a time 1. Press "Add keyword". 2. Type the term (up to 200 characters). Choose a Market: "Project default" or United States, United Kingdom, Ireland, Australia, Canada or New Zealand. Choose a Buying stage, or leave "Not set". 3. Type Monthly volume (a whole number), CPC and Difficulty (0 to 100) if you have them. A blank stays "Not fetched". 4. Press "Add keyword". A term already tracked for the same market is refused with "That keyword is already tracked for this market." The same term in another market is a separate keyword. ### Paste a list 1. Press "Add many". 2. Type or paste one keyword per line. To set a market or stage, append ` | market | stage`, for example `link building services | gb` or `geo agency pricing | us | comparison`. Market codes are `us`, `gb`, `ie`, `au`, `ca` and `nz`. 3. Press "Add keywords". The toast counts "N added · M already tracked · K left in the box to fix", and any row that failed stays in the box with its reason. A pasted list carries no numbers. For a list with volumes, import a CSV. ### Import a CSV 1. Export the list from your keyword tool. Only a keyword column is required. 2. Press "Import CSV" and choose the file (up to 2 MB and 5,000 rows). 3. Read the toast: "N imported · M already tracked · K rows skipped · J rows past the 5,000 limit left out". Columns are read by their usual names, in any letter case: | Column | Header names read | |---|---| | Keyword (required) | keyword, keywords, term, search term, query, search query | | Market | market, country, locale, geo, location | | Stage | intent, stage, buying stage, search intent | | Volume | volume, search volume, monthly searches, searches, avg. monthly searches | | CPC | cpc, cost per click, cpc (usd), cpc usd | | Difficulty | difficulty, kd, keyword difficulty, kd %, seo difficulty | > **Tip:** The market column takes the two-letter code (`gb`, not "United Kingdom"); a row with any other value is skipped, and the toast names the first three skipped rows and why. The stage column takes a stage word such as `purchase` or `problem aware`; anything else leaves the stage "Not set" rather than skipping the row. A file with no header row, or no keyword column, is refused and the message says what to name the column. ## How to link a keyword to its prompts 1. On the Keywords view, press "Link prompts" on the row, or the "N prompts" button once it has links. 2. The popover "Prompts behind “term”" lists the linked prompts, each with how it was linked. Press the × beside a prompt to unlink it. 3. Type in "Find a prompt to link…" and press a prompt to link it. Eight candidates show at a time, and "N more, type to narrow" counts the rest. Each link says how it was made: | Label | Meaning | |---|---| | Linked by hand | You linked it in this popover. | | Prompt written from it | The prompt was written from this keyword. | | Suggested from the prompt | The platform suggested the term from the prompt, and you added it as suggested. | On the Prompts page, a prompt with keywords behind it wears a "N keywords" chip. Open it for "Keywords behind this prompt", each with its market and stage, and "Manage keywords →" to come back here filtered to that prompt. ## Editing and sources Press the pencil on a row to open the inline form. It has the same fields as "Add a keyword", plus a line that says where the keyword came from: "Typed in" or "Imported", and, when its numbers came from somewhere else, where those came from, for example "Imported · numbers typed in". Press "Save" or "Cancel". When you change any of the three numbers, the keyword's numbers are marked as typed in. Leave all three alone and they keep their source. Renaming a keyword to a term and market that match a removed one is refused: "A removed keyword already has this text and market. Restore it from the Archived tab instead." ## Archiving and restoring 1. Press the trash can on a row. The dialog "Remove “term”?" says "It leaves the list and keeps its prompt links. You can restore it from the Archived tab." 2. Press "Remove". The toast reads "Keyword removed. Restore it from the Archived tab." 3. To bring it back, open "Archived" and press "Restore". The toast reads "Keyword restored", and its prompt links return with it. Adding the same term and market again, by hand, pasted or from a CSV, restores the archived keyword instead of creating a twin. A keyword marked "Paused by the plan" has a disabled Restore with the tip "Paused by your plan. It comes back when the plan has room." ## Limits - Up to 500 active keywords per project, on every plan. Archived keywords do not count. An add past the ceiling is refused with "Up to 500 keywords per project. You have N, so there is room for M more; this would add K." A restore counts as an add. - A keyword is up to 200 characters. Volume is a whole number of 0 or more. Difficulty runs 0 to 100. - A CSV is one file up to 2 MB and 5,000 rows. - Markets: the project default plus United States, United Kingdom, Ireland, Australia, Canada and New Zealand. See [Markets and locations](https://www.sourcesignal.ai/docs/tracking/markets-and-locations.md). ## From Spot or your own agent You can work the same list in words. "Find keywords" lists up to 20 at a time with the room left under the 500 ceiling. "Add keywords" takes up to 50 terms with an optional market and stage. "Remove keywords" takes up to 200 and moves them to Archived with their links kept. A change shows a preview first and happens when you press Apply, unless your setting under Settings, Agents applies that kind of change at once, with "Take back" for a day. Apply on a keyword change can be undone for ten minutes. See [Spot, the in-app agent](https://www.sourcesignal.ai/docs/spot-and-agents/spot-the-in-app-agent.md) and [Connect your own agent over MCP](https://www.sourcesignal.ai/docs/spot-and-agents/connect-your-own-agent-over-mcp.md). > **Note:** "Not fetched" is never a zero. The numbers come from your file or your typing; nothing on this page fetches them for you, and nothing here changes a score or a report. ## Related - [Prompts](https://www.sourcesignal.ai/docs/tracking/prompts.md): The Prompts page holds the questions buyers ask AI engines and shows, for each one, how often the answers named you over the last 30 days. Add, import, classify, remove and restore them here. - [Markets and locations](https://www.sourcesignal.ai/docs/tracking/markets-and-locations.md): Choose the countries you sell into, in priority order. The first country is where every prompt is checked; the rest, and any US states you serve, shape how your tracking plan is phrased. - [Spot, the in-app agent](https://www.sourcesignal.ai/docs/spot-and-agents/spot-the-in-app-agent.md): Spot is the panel you open on any page to ask about a project in plain words. Its answers come from your own data, and a change it prepares waits as a card for you to decide, unless your setting applies that kind of change at once. - [Navigation and keyboard shortcuts](https://www.sourcesignal.ai/docs/reference/navigation-and-keyboard-shortcuts.md): Every page shares one frame: a rail of areas, a strip of steps and a search you open with Ctrl K or ⌘ K. Keyboard chords (g, then a letter) jump to any area. --- # Competitors > Build the roster of brands AI names instead of you, compare your share of AI answers with theirs, and see where each one wins and which sources carry it. sourcesignal documentation, Tracking. Updated 2026-10-18. The page for people: https://www.sourcesignal.ai/docs/tracking/competitors. Every article: https://www.sourcesignal.ai/llms.txt The Competitors page is where you decide which brands count as competition and read how often AI answers name them instead of you. It sits beside [Prompts](https://www.sourcesignal.ai/docs/tracking/prompts.md) and [Keywords](https://www.sourcesignal.ai/docs/tracking/keywords.md) in your project's navigation. Open it when a new brand keeps turning up in your answers, or when a competitor's name or website needs correcting. The Roster and Mentions & citations views count the AI answers stored for your tracked prompts over the last 30 days, the newest 5,000 at most. The Visibility view is different: it shows Ahrefs Brand Radar's own numbers for the brands on your report. ## What you see Three views run across the top: "Roster", "Visibility" and "Mentions & citations". The "Re-read site" control sits at the end of the same line. ### Your share of AI answers On the Roster and Mentions & citations views, one band leads the page: "Your share of AI answers · last 30 days". It shows your share as a large percentage, a trend chip ("↑ 2.3 pts", "↓ 1.0 pts" or "→ flat"), "Named in N of M answers" and one sentence on where you stand, such as "3 competitors are named more often than you." Beside it, bars draw you and the five most named competitors on one scale. If the band asks you to add your brand name, open Settings, then Project details, and fill in "Brand name". ### Brands seen in your answers Under the band, the Roster view opens with the suggestions queue, "Brands seen in your answers · N not tracked": brands found in your AI answers that are not on the roster yet. Each row shows the brand and its evidence, "N prompts · N answers", counted since monitoring began. The six companies with the most evidence come first, then "Show N more · M not shown". Names made only of category words, such as "GEO Agency", sit apart in a folded line, "N category phrases held back until their site is cited". Such a name joins the roster only once its own site is cited in your AI answers. At the foot of the queue sits the rule "Add brands automatically once named in N or more prompts". When the queue is empty, the line in its place says why. ### The roster "Roster · N tracked" lists who AI names in your answers, with a "Find a competitor…" box and a sort ("Share of answers", "Biggest movers up", "Newest first", "Name A–Z"). | Column | What it shows | |---|---| | Competitor | Logo, name and domain. A "category name" tag marks a name made of category words. | | Share of answers | How often your stored answers name the brand over the last 30 days, as a percentage and a bar. | | Trend | How the share moved across the 30 days, in percentage points, newer half against older half. Small moves show as "→ flat". | | Where they win | The prompt where AI most often names this competitor and not you, with "N answers without you". Otherwise "Only named alongside you" or "Not named in the last 30 days". | | Sources AI cites for them | The third-party sites AI cites most in answers naming this competitor, leaving out every competitor's site and yours. | | Added | The date the brand joined your roster. | Under the table, "Phrase uses left out · Acme −12 · N answers in the last 30 days" counts the answers every figure on the page leaves out for that competitor. ### Visibility This view fills in once a Brand Radar report is connected and imported. Until then it reads "No visibility data yet." A lead line names the brand with the highest AI share of voice beside yours, with the date of the import. One row per brand, yours marked "you": Mentions, Citations, Impressions, AI Share of Voice and a 30-day "SoV trend" sparkline. The first three are sums over the last 30 days; AI Share of Voice is the average of Ahrefs' daily readings. A dash means nothing was recorded, not zero. Category-phrase names are listed apart under "Held out · category-phrase names". ### Mentions & citations The stored answers, broken down per competitor: - "Named in responses": answers from the last 30 days whose text names the competitor. - "Cited in responses": answers that cite one of its pages, named or not. - "Prompts named in": prompts with at least one answer naming it. - "Cited pages": every page from its site in your cited pages, all time. A "+" marks a floor. - "Sample prompts": a few of the questions where it was named. A competitor with nothing to show reads "Tracked, not observed in your response data". ### The detail sheet Click any row, name or bar to open the competitor's sheet: "Last 30 days" tiles for the four Ahrefs numbers and a "Daily timeline" of them, "AI answers naming or citing this competitor (N)" with excerpts (engine, date, prompt and "Open in snapshot →"), and "Pages from this domain cited by AI (N)", an all-time list with "Open →" into [Opportunities](https://www.sourcesignal.ai/docs/opportunities/what-an-opportunity-is.md). ## How to build the roster from the queue 1. Open Competitors; the queue sits at the top of the Roster view. 2. Press "Add" on each real competitor. Its website is worked out for you; if it cannot be, "Couldn't work out a domain for “Name”. Add it manually." opens a site box with "Save". 3. Press "Dismiss" (the cross) on names that are not competitors. A dismissed brand is not suggested again. 4. Press "Add all" to take the queue's companies in one press (it reads "Add top 50" past 50 companies). Brands whose site cannot be worked out stay in the list. 5. For a category phrase that is a real company, press "Site" and enter its domain. Until that site is cited in your AI answers, the add is refused with the reason and the phrase stays in the queue. 6. Check "Add brands automatically once named in N or more prompts" and set the number (default 3). Switching it on runs a pass at once; after that, qualifying brands join after each data update. Each needs a website that can be worked out; the rest stay in the queue. ## How to correct a competitor's details 1. Find the row with "Find a competitor…". 2. Press "Edit" (the pencil) and change the Name or Domain. 3. Press "Save changes". The toast "Competitor updated" confirms it, and the brand's recorded history follows the new name. 4. To take a brand off the roster, press "Remove" (the bin), then "Remove" again under "Remove Acme from the cohort?". Its history is kept for reporting, and re-adding the brand later restores it with that history. ## Limits - Shares, trends, wins and sources count the stored answers of the last 30 days, the newest 5,000 at most. Cited pages count all time. - "Add all" acts on the 50 companies with the most evidence per press. The automatic rule takes a whole number of 1 or more. - A site read adds up to 100 prompts and 15 competitors per go. "Re-read homepage" is on every plan. "Deep dive" comes with Growth, Agency and Enterprise, and on a trial it waits for a paid plan ([Plans and limits](https://www.sourcesignal.ai/docs/account-and-billing/plans-and-limits.md)). ## Good to know > **Note:** The Visibility view and the roster can disagree. Visibility is Ahrefs Brand Radar's tracking of its own answer set; the roster counts the answers stored here, and only those whose text names the brand. > **Tip:** "Sources AI cites for them" is the first place to look for a placement of your own: the sites AI already cites on the questions where a competitor is named. The link is an association with its mentions, not proof they caused them. - "Re-read site", then "Re-read homepage" or "Deep dive", proposes competitors (and prompts) from your own pages; you press "Add selected" on the ones you keep. An empty roster with nothing waiting in the queue also offers "Add one by hand", which opens the competitors section of the setup page. - AI platforms and big-tech properties (ChatGPT, Google, YouTube and the like) never join the roster. If one truly competes with you, edit an existing competitor's name and domain to it. - A removed brand may be suggested again if it keeps appearing in your answers. - Suggested brands also reach you as "New competitor" cards on the Autopilot deck, where "Track competitor" adds them ([The deck](https://www.sourcesignal.ai/docs/autopilot/the-deck.md)). Spot can list the roster and the queue and add up to 10 competitors at a time, previewed before you press Apply ([Spot, the in-app agent](https://www.sourcesignal.ai/docs/spot-and-agents/spot-the-in-app-agent.md)). ## Related - [Prompts](https://www.sourcesignal.ai/docs/tracking/prompts.md): The Prompts page holds the questions buyers ask AI engines and shows, for each one, how often the answers named you over the last 30 days. Add, import, classify, remove and restore them here. - [Topics and competitors in the report](https://www.sourcesignal.ai/docs/reports/topics-and-competitors.md): The Topics and Competitors tabs show where your AI visibility rises or falls, subject by subject, and how your share of AI voice compares with a fixed set of rival brands. - [The deck](https://www.sourcesignal.ai/docs/autopilot/the-deck.md): Your deck deals the decisions only you can make, one card at a time. This article covers every card, the keys, undo, the daily budget and what a cleared deck means. - [What an opportunity is](https://www.sourcesignal.ai/docs/opportunities/what-an-opportunity-is.md): An opportunity is a page an AI engine cited while answering one of your tracked prompts. Each page gets one record that carries its type, its 0 to 100 priority, when and where it was cited, and whether it already names you. - [The answer feed and data updates](https://www.sourcesignal.ai/docs/tracking/the-answer-feed-and-data-updates.md): Connect one Ahrefs Brand Radar report, managed for you or your own, and the daily data update brings in AI answers, cited pages and visibility every morning. The Data tab keeps every run, the full schedule and the way to fix a feed that stops. --- # Markets and locations > Choose the countries you sell into, in priority order. The first country is where every prompt is checked; the rest, and any US states you serve, shape how your tracking plan is phrased. sourcesignal documentation, Tracking. Updated 2026-10-18. The page for people: https://www.sourcesignal.ai/docs/tracking/markets-and-locations. Every article: https://www.sourcesignal.ai/llms.txt Target markets tell sourcesignal where your buyers are. You pick them when you create a project and can change them any time in the project's settings. The first country on your list is where every prompt is actually checked, so it decides which AI answers, competitors and cited pages you are measured against. Your markets, in the order you list them, and the US states you serve shape how the tracking plan phrases buyer questions and which competitors it proposes. You come back to this screen when you expand into a new country, when you start serving new states, or when the Ahrefs Brand Radar card warns that prompts are filed under the wrong country. ## Where to find it - When you create a project (the "Your brand" step of the setup wizard), the form carries the "Target markets" chooser. Creation needs at least one market: "Pick at least one target market. The first decides which country every prompt is checked in." - Afterwards, open the project's Settings. "Project details" is the first card, with "Project name", "Brand name", "Industry or niche (optional)" and "Target markets". Any member who can open the project can edit it. - During setup, the "Your brand" card (on the setup page, and on the wizard's "Your brand" step when you return to it) holds the same form behind "Edit". It also needs at least one market. ## What you see The **Target markets** chooser is an ordered list, not a drop-down: - A counter reading "None chosen yet" or "N chosen · first is primary". - A ranked list of the markets you chose, each with its flag. On a wider screen, the first country carries the note "prompts are checked here" and Worldwide carries "the plan is not shaped to one country". - On every row, arrows for "Higher priority" and "Lower priority" and an X to "Remove". - An "Add:" row of chips for the markets not yet on the list: Worldwide, United States, United Kingdom, Ireland, Australia, Canada and New Zealand. - The helper line: "In priority order. The first country is where every prompt is actually checked. It changes which AI answers, competitors and cited pages you are measured against. The rest shape the tracking plan." **States you serve** appears only while the United States is on the list: - A counter reading "None picked · nationwide" or "N chosen · in priority order", and the line "Nationwide (all US states)" when none are picked. - An "Add states" disclosure with a "Find a state…" box and a chip for each remaining state ("No state matches." when the search finds nothing). The list covers the 50 states and the District of Columbia. - The helper line: "Shapes how buyer questions are phrased (“best … in Texas”) and which competitors are proposed, in this order. Prompts are still checked country-wide." An older project may show a warning instead of a list: "Saved as “EMEA”, which is not a market we check. Prompts fall back to the US until you choose one below." Pick a market from the chips and save to clear it. ## What the first country changes The first real country on your list is the project's primary market. Three things follow from it: - **The AI answers.** Your prompts are asked as a searcher located in that country. When you open a captured answer, the line under its title shows the date, the engine and, when recorded, the country. - **The answer feed.** Ahrefs Brand Radar files each prompt under one country. On a report the platform manages for you, the Ahrefs Brand Radar card (the Visibility tracking section of Settings) says which one: "Prompts written here sync under GB (this project’s target market, United Kingdom)." With no market set, the card says prompts sync under US "because this project has no target market set" and asks you to set one above. See [The answer feed and data updates](https://www.sourcesignal.ai/docs/tracking/the-answer-feed-and-data-updates.md). - **Competitors and cited pages.** The brands your [competitor roster](https://www.sourcesignal.ai/docs/tracking/competitors.md) suggests, and the pages that become opportunities, come from those answers. Change the country and you are measured against a different set of both. > **Note:** Worldwide is a market like any other on the list, but it cannot be the check country. It tells the plan not to phrase questions for one country; the first real country on the list still decides where prompts are checked. A project with only Worldwide chosen is checked from the United States. ## What your markets and states shape Your markets, in the order you list them, and the states you serve shape the buyer questions and competitors that the setup wizard and the site read propose. The wizard reads your site, maps the service areas it offers and drafts a group of questions for each one, phrased for your markets in priority order. States shape the wording and the competitors proposed, so a business serving Texas can get questions like "best … in Texas". ### Local questions A local question gets its place from its own words. The plan can phrase its drafts for the states you serve, and you can type a city or state into any question you add on the [Prompts](https://www.sourcesignal.ai/docs/tracking/prompts.md) page. That question is still asked as a searcher in the primary country. No setting pins a prompt to a state or city: prompts are checked country-wide. ## How to change your markets 1. Open the project's Settings and find the "Project details" card. 2. Under "Target markets", press an "Add:" chip for each market you sell into. 3. Order them with the arrows ("Higher priority" and "Lower priority"). The first country is where prompts are checked. 4. Remove a market with its X. 5. If the United States is on the list, open "Add states", search with "Find a state…", press the states you serve and order them the same way. Leave the list empty for nationwide. 6. Press "Save changes". You see "Project updated". A refusal appears under the form and as a toast: "Unknown market", "Each market once" or "Unknown state code". Removing the United States drops the states you picked when you save; add it back to pick them again. ## After a change: re-file your prompts Prompts you already track stay tracked. On a platform-managed Brand Radar report, prompts that sit under the old country are flagged on the Ahrefs Brand Radar card. > **Heads up:** The warning reads "3 prompts in this report are filed under GB, not US. They are being checked against the wrong country and will not show in a US view. Run Sync prompts now to re-file them, then delete the old copies in Ahrefs." Press "Sync prompts now"; the toast confirms, for example "1 prompt had been filed under another location: re-added under US; remove the old copies in Ahrefs". ## Markets elsewhere in the app - **Projects page and switcher.** Each project card reads brand name · industry · primary market, and the switcher tip shows the same line. - **Setup page.** The "Your brand" card lists your markets with the states in brackets, for example "United States (Texas, California)". - **Opportunities.** Once pages have been located, the "Filters" panel has a "Location" group listing the locations found among your cited pages: "Global / no market", the six countries, "Europe", "North America", "Other market" or "Not classified yet". "Match this project’s markets (United Kingdom, United States)" checks the ones that fit your markets, plus pages that name no market and "Not classified yet". The detail panel's "Location" line names the market, the place when known, and how it was read. See [The queue, views and filters](https://www.sourcesignal.ai/docs/opportunities/the-queue-views-and-filters.md). - **Keywords.** Each keyword is tracked in a market of its own: "Project default" or one of the six countries. A chip shows it only when it differs from the project default. See [Keywords](https://www.sourcesignal.ai/docs/tracking/keywords.md). ## Limits | What | Limit | |---|---| | Market options | 7 (Worldwide plus six countries), each at most once | | States you serve | 51 (the 50 states and the District of Columbia), each once, only while the United States is listed | | Primary markets per project | 1 on Brand, Growth and Agency; custom on Enterprise | You can list as many of the seven markets as you sell into, but prompts are checked in one country per project. The [pricing table](https://www.sourcesignal.ai/pricing) shows this as "Primary markets". For the rest of your plan's limits, see [Plans and limits](https://www.sourcesignal.ai/docs/account-and-billing/plans-and-limits.md). ## Good to know - "Industry or niche (optional)" sits on the same card and is "Context for AI summaries and briefings. Does not affect what is checked." - The [free AI visibility report](https://www.sourcesignal.ai/free-report) offers the same seven markets, and a report on Worldwide is also asked from the United States. - The markets you choose during setup feed the proposals you approve before anything is added; see [Set up your first project](https://www.sourcesignal.ai/docs/getting-started/set-up-your-first-project.md). ## Related - [Prompts](https://www.sourcesignal.ai/docs/tracking/prompts.md): The Prompts page holds the questions buyers ask AI engines and shows, for each one, how often the answers named you over the last 30 days. Add, import, classify, remove and restore them here. - [The answer feed and data updates](https://www.sourcesignal.ai/docs/tracking/the-answer-feed-and-data-updates.md): Connect one Ahrefs Brand Radar report, managed for you or your own, and the daily data update brings in AI answers, cited pages and visibility every morning. The Data tab keeps every run, the full schedule and the way to fix a feed that stops. - [Competitors](https://www.sourcesignal.ai/docs/tracking/competitors.md): Build the roster of brands AI names instead of you, compare your share of AI answers with theirs, and see where each one wins and which sources carry it. - [The queue, views and filters](https://www.sourcesignal.ai/docs/opportunities/the-queue-views-and-filters.md): The Opportunities table lists every page AI engines cite for your prompts, ordered by priority. Narrow it with filters, quick chips and saved views, sort it, export it, and follow a prompt straight to the pages it cites. - [Plans and limits](https://www.sourcesignal.ai/docs/account-and-billing/plans-and-limits.md): What Brand, Growth, Agency and Enterprise include, how each limit counts and what happens when you reach one. When a plan is smaller than what you track, nothing is deleted: you choose what stays and the rest is paused. --- # AI engines > Each project tracks its buyer questions on its own set of AI engines, chosen within your plan's count. This article covers the catalog, the AI engines card, how often each engine is checked (its cadence), paused engines and the engine status in the rail foot. sourcesignal documentation, Tracking. Updated 2026-10-18. The page for people: https://www.sourcesignal.ai/docs/tracking/ai-engines. Every article: https://www.sourcesignal.ai/llms.txt Every project tracks its buyer questions on its own set of AI engines. You choose that set under Project settings, then AI engines, within the number your plan allows. The same card shows what your connected answer feed is running right now, so a save never changes your data quietly. You come here when you set up a project, when you change plan, or when the schedule says an engine is off. The rail foot on every project page shows what the platform is doing for the project at this moment, including the checks that collect from these engines. ## Which engines you can track The catalog holds eight answer engines: ChatGPT, Gemini, AI Overviews, AI Mode, Claude, Perplexity, Copilot and Grok. A few further engines appear on the grid as not available. A card that nothing can collect shows its reason in place of its sentence, and it cannot be switched on. The reasons: your Ahrefs feed is not connected yet (one of the eight, on a project with no feed), no collection route exists for it yet, it is not available for our weekly sample yet, or it is an errand-running agent rather than an answer surface and not comparable with the rest. Each project picks its own set. The card's lead says so: "Choose the answer engines this project tracks. Each project picks its own set. They are not shared across your projects." Nine engines on Agency means nine on every project, not nine shared between them. ## How many each plan allows | Plan | AI engines per project | |---|---| | Brand | 3 | | Growth | 6 | | Agency | 9 | | Enterprise | Custom | A trial runs on Growth's limits, so it tracks up to 6 per project. The count sits at the top right of the card: "4 of 6 on your plan", or "4 on · your plan sets no limit". Engines over the count are paused, never deleted. The full table of limits is in [Plans and limits](https://www.sourcesignal.ai/docs/account-and-billing/plans-and-limits.md). ## What you see on the card - A grid of toggle cards, one per engine: logo, name, a check when selected, and one sentence about what this project does with it. - Small status lines under the sentence when they apply: "Paused: over your plan", and the report's live state. A selected engine the report runs reads "Running in your feed". Switch it off and the line turns to warning color: "Saving switches this off" on a report the platform manages, or "Still on in your report" on a report you own. - The info hint "What each card says" explains both lines: the sentence names the route the engine is collected by (your own Ahrefs feed, our weekly sample, or both); the small line is your Ahrefs report as read back on the last update. - Notices above the grid, when they apply. At the count, owners and admins see: "You are using every engine the Growth plan includes. Agency tracks more." with an "Upgrade" button that opens Billing. Over it: "2 engines paused. Brand includes 3, and you have more selected. They are kept, not deleted. Drop one below, or upgrade, and they start collecting again." On a report you own that runs more than you collect: "Your own Brand Radar report is running 2 engines this project does not collect. We never change a report you manage yourself." Before a first read-back: "We have not read your feed's own engine list yet. It is read back on the next scheduled update." - The button "Save engines", or "Save this selection" before a first save, with "Unsaved changes" beside it once you have tapped a card. ### The sentence under each engine | Sentence | Meaning | |---|---| | collected via your Ahrefs feed | Your report runs it today. | | set up to collect via your Ahrefs feed | Set up with it; the report has not been read back yet. | | would be added to your Ahrefs feed | Off. Switching it on adds it to a report the platform manages. | | will be added to your Ahrefs feed on save | Switched on, not saved yet. | | being added to your Ahrefs feed | Saved; the next update confirms it. | | would need adding to your own Ahrefs report | Off, on a report you own. You add it in Ahrefs. | | waiting for your own Ahrefs report to run it | Chosen here, not yet running in your own report. | | collected in our weekly sample | The weekly sample asks this engine for this project. | | joins our weekly sample on save, leaves our weekly sample on save, would join our weekly sample | What a save would do to the sample. | | collects when you drop another engine or upgrade | Paused by your plan's count. | When both routes cover an engine, " + our weekly sample" is appended. ## How to choose a project's engines 1. Open Project settings, then AI engines, and read the count and each card's two lines. 2. Tap cards on or off. Save stays off until you change something. 3. Stay within the count. Over it, Save is disabled and the note says how many to remove: "Growth tracks 6 engines per project. Remove 2 to save." 4. Press "Save engines". The toast reads "Engines saved." When the change could not reach your feed, the toast says so instead, for example: "Saved, but Ahrefs did not accept the change (…). Your next daily update will try again." Before a first save the note reads "This is what the project collects today. Change one to make it your choice." Nothing changes until you tap a card and save. > **Heads up:** On a report the platform manages, switching off an engine the feed is running stops its answers arriving for this project once you save. The card says so before you press Save: "Saving switches this off". ## Two routes to the answers Your answer feed is an Ahrefs Brand Radar report. For a report the platform created for you, your choice is written into the report on save, and the daily data update keeps the two in line. For a report you connected yourself, nothing is ever written: you turn engines on and off in your Ahrefs account, and the card only reports what it runs. See [The answer feed and data updates](https://www.sourcesignal.ai/docs/tracking/the-answer-feed-and-data-updates.md). The weekly sample is the platform's own run of the AI engines for your tracked prompts. On the schedule it is the row "AI answers run", Mondays at 07:30 UTC. It runs alongside your feed and is never mixed into the same line. Only enrolled projects have it; others read "not enrolled". ## Cadence per engine You choose which engines, not how often. The cadence belongs to the report and is shown, not edited. It appears under the "Data update" row of the Schedule tab (Project settings, then Data), on the console's Schedule card, and on the Pipeline tab, where "Collect fresh answers from" lists the engines grouped by cadence. - Platform-managed report: "ChatGPT, Gemini weekly · Perplexity daily. You pick these under AI engines in project settings." With every engine off: "No engines are switched on. Nothing new will arrive until you turn one on under AI engines in project settings." - Your own report: the same clauses, then "Set in your own Ahrefs report, not here." - Over the plan: "2 engines in the report are not on your plan." followed by what switches them off: the next update, picking this project's engines, or turning them off in your own report. A report the platform creates starts with ChatGPT, AI Overviews and Perplexity, weekly, unless the project already has a saved choice, which it then follows. An engine you switch on runs weekly; one already running keeps the cadence it had. The line is absent until the engine list has been read back on a scheduled update. ## Paused engines An engine is paused when your selection is over the plan's count, usually after a move to a smaller plan. It is kept, not deleted, and nothing collects it. Its card reads "Paused: over your plan" and "collects when you drop another engine or upgrade". Save is refused while the selection is over the count. The What stays picker in Billing covers projects and prompts, not engines. Engines pause on each project by themselves, and the card marks which ones. To bring one back, drop another engine and save, or upgrade. An upgrade restores them without retyping anything. ## Engine status in the rail foot The foot of the left rail (the bottom bar on a phone) shows what the platform is doing for this project right now. Spot sits beside two short lines. - Working: the task's name over what it is doing, for example "Checking for new data" over "Checking the feed for new answers", or "AI pipeline" over its current step. Other names you will see: "Daily upkeep", "Syncing prompts to Ahrefs", "First-party sampling" (the weekly sample), "Finding contacts", "Writing a page". When more runs at once, the tip adds how many ("2 more running"). A thin progress strip along the top edge shows how far along when a real ratio exists. - A warning mark appears when nothing has moved for a while; its tip says how long and what it may be waiting on. - After a task you started finishes, the result is held: "AI pipeline finished" followed by a count of what it enriched, "Promoting to outreach finished · 3 drafted · 1 need a contact", "… stopped" with the error, or "… canceled". Finished results stay about a minute; failures stay until you dismiss them with the small X ("Dismiss this result"). - Quiet: "All quiet", with Spot snoozing. - "Engine room" opens the drawer where the work in progress and held results appear in full. See [The engine room and pipeline](https://www.sourcesignal.ai/docs/autopilot/the-engine-room-and-pipeline.md). With the rail collapsed, Spot alone carries the whole sentence as its tip. On a phone, the sentence runs across the bottom bar while anything is running. ## Good to know - Owners and admins change engines. Members see the same cards and "Changing engines needs admin access." - During setup, the Tracking step says "The engines are chosen per project." and its info hint "Engines" adds "Change them any time in the project's settings, under AI engines." - If the selection cannot be read: "Engine selection is still switching on for this workspace. Your data keeps arriving on the engines it already uses." - Ask Spot, or your own agent over MCP, for "Engine mention rates" (how often each engine names you over a window of complete days, with a margin) and "What engines searched" (the searches the engines ran while answering your prompts). Both are reads; nothing changes. See [Spot, the in-app agent](https://www.sourcesignal.ai/docs/spot-and-agents/spot-the-in-app-agent.md). ## Related - [The answer feed and data updates](https://www.sourcesignal.ai/docs/tracking/the-answer-feed-and-data-updates.md): Connect one Ahrefs Brand Radar report, managed for you or your own, and the daily data update brings in AI answers, cited pages and visibility every morning. The Data tab keeps every run, the full schedule and the way to fix a feed that stops. - [Plans and limits](https://www.sourcesignal.ai/docs/account-and-billing/plans-and-limits.md): What Brand, Growth, Agency and Enterprise include, how each limit counts and what happens when you reach one. When a plan is smaller than what you track, nothing is deleted: you choose what stays and the rest is paused. - [The engine room and pipeline](https://www.sourcesignal.ai/docs/autopilot/the-engine-room-and-pipeline.md): Where you watch the autopilot work: the engine room shows the run in progress and your levers, and the Pipeline page shows the steps your data moves through, each with a status and when it moves next. - [Prompts](https://www.sourcesignal.ai/docs/tracking/prompts.md): The Prompts page holds the questions buyers ask AI engines and shows, for each one, how often the answers named you over the last 30 days. Add, import, classify, remove and restore them here. - [Set up your first project](https://www.sourcesignal.ai/docs/getting-started/set-up-your-first-project.md): The six-step wizard turns one brand into a running project: your brand, research, competitors, prompts, tracking and mailbox. You can leave part way through, and the setup companion brings you back. --- # The answer feed and data updates > Connect one Ahrefs Brand Radar report, managed for you or your own, and the daily data update brings in AI answers, cited pages and visibility every morning. The Data tab keeps every run, the full schedule and the way to fix a feed that stops. sourcesignal documentation, Tracking. Updated 2026-10-22. The page for people: https://www.sourcesignal.ai/docs/tracking/the-answer-feed-and-data-updates. Every article: https://www.sourcesignal.ai/llms.txt Your project reads its AI answers from one answer feed: an Ahrefs Brand Radar report. sourcesignal can create and manage that report for you with one click, or you connect a report you already own. Once a feed is connected, a data update runs every morning at 06:00 UTC. It collects the latest AI answers, the pages they cite and your brand's visibility, and whatever is new moves through the processing steps without you touching anything. The Data step of the project's Settings area is where you watch that happen. It opens on three tabs: "Imports" (one row per import run), "Snapshots" (the datasets each update saved, plus any you upload; see [Snapshots](https://www.sourcesignal.ai/docs/tracking/snapshots.md)) and "Schedule" (the timetable of everything that runs for the project). ## Connect the answer feed Open Project settings, then "Visibility tracking". The card is titled "Ahrefs Brand Radar". Leaving the report blank switches the daily update off. ### One-click setup Where it is offered, the box "One-click setup" creates a Brand Radar report for the project and keeps its prompt list in sync: prompts you add or remove here follow into the report. 1. Type your brand's canonical domain in "Brand domain" (a domain, not a display name). The button stays disabled with "Enter the brand domain first." until you do. 2. Press "Create & connect report". 3. The box now reads "Platform-managed report" and names the country your prompts are checked in, taken from the project's target market (see [Markets and locations](https://www.sourcesignal.ai/docs/tracking/markets-and-locations.md)). "Sync prompts now" pushes prompt changes to the report straight away. If the card warns that prompts are filed under another country, it re-files them; then delete the old copies in Ahrefs. > **Heads up:** Changing the report ID under a platform-managed report hands management back to you. sourcesignal stops writing to it. ### A Brand Radar report you already own 1. Optional: an admin saves your organization's Ahrefs API token under Settings, then Organization, in the card "Ahrefs API token". It lets you browse your reports and is never shown back. 2. On the "Ahrefs Brand Radar" card, paste the report ID or the whole Brand Radar URL into "Brand Radar report". With a token saved, you can press "Browse my reports" and pick one from "Choose a Brand Radar report" instead. 3. Press "Save". 4. Enter the "Brand domain" and press "Save". Without it, the mention, share-of-voice and impressions history is not collected. A report you own is read, never written to: engines are turned on and off in your own Ahrefs account (see [AI engines](https://www.sourcesignal.ai/docs/tracking/ai-engines.md)). ## The Imports tab An import is one data update as a record: one morning's answers, cited pages and visibility. There is one import per project per UTC day. The header shows "Last update" with its time, a live countdown ("Next data update 06:00 UTC · in 15 h") and the button "Run import now". Each row leads with the run day and "Scheduled" or "Manual". The columns: - **Run**: how the run ended: Ready, Failed (it can be partial), Nothing new, Running or Queued. - **Datasets**: how many of the three arrived, as "2/3". They are cited pages, brand visibility and raw AI answers. Fewer than 3 is not the same as zero. - **Pages** and **Answers**: what this run saved for its date, not the project total. - **By engine**: cited pages each engine contributed. A page cited by two engines counts under both, so these can exceed Pages. A failed row prints the first thing it is missing in plain words. The trash icon deletes an import and its child snapshots permanently once you confirm "Delete"; "Older imports →" pages back through the history. ## An import's detail Click a row to open the run. The subtitle gives the status, "Scheduled run for" or "Manual run for" the project, how long it ran and what it saved. A notes card follows the run's tone. Failed: "Import errors" lists every line in plain words, with "Retry" (or "Open settings") and "Contact support". Finished with gaps: "What this import is missing · N" lists each gap in plain words, then "Everything else came in." with "Show them" for the feed's notes on how the rest was collected. Three dataset cards, "Cited pages", "Visibility" and "AI answers", show their row counts and an "Open" link to the snapshot. A dataset that did not arrive carries the badge "Missing" and "This dataset wasn't imported. Re-run from the header to retry." ## How to run a check now A manual run collects whatever the engines have produced since the last run, and nothing more. 1. On the Imports tab press "Run import now". It is disabled with "Connect your data feed first." when the project has no feed. 2. In "Run an import now", confirm the project and press "Run import". If today's import already exists, it is replaced. 3. Read the toast ("Import complete.", "Import partially succeeded: some datasets are missing." or "Import finished with errors. Check the row for details.") and open the new row. "Run import now" on a failed row, or on an import's page, starts a fresh import for today. It never retries the old run and cannot re-run a historical day. While a new project waits for its first answers, the console's first-run card ("Your questions are queued with the engines") offers "Check now" instead. Checks have a short cooldown: three minutes after the last run, one minute after a failed one. When answers land, the toast says how many answers landed, for example "N answers landed. The console is waking up." ## Nothing new A data update that asks the feed and finds nothing newer than the answers already held ends as "Nothing new". No data was pulled and no AI work ran. It is a healthy outcome, shown muted and never red. The run's page reads "Nothing new this run", and the Schedule's Note reads "nothing new yet". ## The Schedule tab The Schedule tab is the full timetable of everything that runs for the project, on a UTC clock. It shows every row, and a muted row's Note says what it waits for ("connect your data feed", "connect GA4 or Search Console", "no verified wins yet", "connect a mailbox", "project paused" or "project archived"). | What runs | Cadence | Applies when | |---|---|---| | Data update | Daily 06:00 UTC | A feed is connected | | First answers | Hourly at :15 UTC until your first answers land | The first two days of tracking | | Traffic refresh | Daily 06:30 UTC | Google Analytics 4 or Search Console is connected | | Placement check | Daily 07:30 UTC | The project has verified wins | | Weekly briefing | Mondays 07:00 UTC | Always | | Outreach sends | Every 5 min inside your send window | A mailbox is connected | Under "Data update" one line names the engines the feed runs and how often, read back from the report on each scheduled update: "ChatGPT, Gemini weekly · Perplexity daily. You pick these under AI engines in project settings." For a report you own it ends "Set in your own Ahrefs report, not here." You choose which engines; the cadence belongs to the report. Every row's info hint ends with your own clock ("06:00 UTC is 07:00 where you are"). "See the full pipeline →" opens the Pipeline tab (see [The engine room and pipeline](https://www.sourcesignal.ai/docs/autopilot/the-engine-room-and-pipeline.md)). ## When the feed fails When a data update fails, a notice appears in the chip at the bottom right of the app. Its title names the state and the project: "Tracking is down" (the update failed), "Tracking is paused" (every engine in the report is off) or "Tracking isn't connected" (no report is set). Its line says what to do. On a wider screen the chip carries the buttons on that project's pages, and "Open project" elsewhere. Its title always opens the project's Data tab. The failed import's "Import errors" card carries the same buttons. A failure on the very first check after connecting is treated as the feed warming up, not as an outage. The notice clears on its own once a run reaches the feed again. ### How to fix a failing feed 1. If the notice names a fix only you can make, make it first: turn an engine on in your own Ahrefs report, or reconnect the Ahrefs token under Settings, then Organization. 2. Press "Retry". When a platform-managed report is missing or has every engine off, it reconnects the report and re-sends your prompts; otherwise it runs the import again. A good outcome reads "Reconnected. The import landed." or "The import landed." 3. Where the notice shows "Open settings" instead (your own report no longer exists, or none is connected), press it and choose a report on the "Ahrefs Brand Radar" card. 4. If it keeps failing, press "Contact support". The support drawer opens with the failing import, the report and how long it has been failing attached (see [Support](https://www.sourcesignal.ai/docs/trust/support.md)). 5. Confirm on the Imports tab that the newest run reads Ready or Nothing new. ## From Spot or your own agent Ask for "Tracking health" when a quiet stretch worries you: it reads the feed's checks and the answers stored by day over 7 to 30 complete days and says whether a dip was an outage or a quiet feed. "Find jobs" lists the project's runs of the last 30 days, 50 at most, and "Run a job again" prepares a fresh run of a failed or canceled one. A prepared run shows as a change card and starts when you press Apply, unless your setting under Settings, Agents applies that kind of change at once. See [Spot, the in-app agent](https://www.sourcesignal.ai/docs/spot-and-agents/spot-the-in-app-agent.md) and [Connect your own agent over MCP](https://www.sourcesignal.ai/docs/spot-and-agents/connect-your-own-agent-over-mcp.md). ## Related - [AI engines](https://www.sourcesignal.ai/docs/tracking/ai-engines.md): Each project tracks its buyer questions on its own set of AI engines, chosen within your plan's count. This article covers the catalog, the AI engines card, how often each engine is checked (its cadence), paused engines and the engine status in the rail foot. - [Snapshots](https://www.sourcesignal.ai/docs/tracking/snapshots.md): Each data update saves its AI answers, visibility figures and cited pages as snapshots, one dataset saved on one day. Open one, compare two visibility snapshots, upload a Brand Radar CSV by hand, or delete a snapshot you no longer need. - [The engine room and pipeline](https://www.sourcesignal.ai/docs/autopilot/the-engine-room-and-pipeline.md): Where you watch the autopilot work: the engine room shows the run in progress and your levers, and the Pipeline page shows the steps your data moves through, each with a status and when it moves next. - [Markets and locations](https://www.sourcesignal.ai/docs/tracking/markets-and-locations.md): Choose the countries you sell into, in priority order. The first country is where every prompt is checked; the rest, and any US states you serve, shape how your tracking plan is phrased. - [Support](https://www.sourcesignal.ai/docs/trust/support.md): Support is a conversation with the sourcesignal team that lives in the app and in your inbox. Here is how to reach it, what to send, and what to do when a screen in the app does not load. --- # Snapshots > Each data update saves its AI answers, visibility figures and cited pages as snapshots, one dataset saved on one day. Open one, compare two visibility snapshots, upload a Brand Radar CSV by hand, or delete a snapshot you no longer need. sourcesignal documentation, Tracking. Updated 2026-10-18. The page for people: https://www.sourcesignal.ai/docs/tracking/snapshots. Every article: https://www.sourcesignal.ai/llms.txt Each data update saves up to three datasets for your project: the AI answers it collected, your brand's visibility figures and the pages those answers cited. Each is kept as a snapshot: one dataset, saved on one day. The Snapshots tab lists all of them, together with any CSV export you upload yourself. You open it to read the raw answers behind a change, check each brand's figures for one period, compare two periods, or bring a Brand Radar export in by hand. In the left rail, open Settings, choose Data, then the "Snapshots" tab (Imports is the default, Schedule sits beside it). The imports themselves are covered in [The answer feed and data updates](https://www.sourcesignal.ai/docs/tracking/the-answer-feed-and-data-updates.md). ## What you see The header reads "Snapshots" with two buttons, "Compare snapshots" and "Upload snapshot". A filter bar sits above the list: "Search snapshots" matches a label or a date typed as `2026-09-23`, "Dataset" narrows to "Cited pages", "Visibility" or "AI answers", and "Uploaded from (UTC)" and "Uploaded to (UTC)" bound the day a snapshot was saved. "Clear filters" appears once any filter is set. Filters apply to the whole project, not only the rows on screen, and a status line counts the snapshots shown against the total, such as "50 of N snapshots". Each row shows a dataset badge, a title (the day the snapshot covers, "Wed 23 Sep 2026", or the period label you typed on upload), when it was saved ("uploaded 3 days ago"), its size ("N cited pages"), and "N import issues" when rows were skipped. Click the title to open it; the trash icon deletes it. The list shows the newest 50, with "Older snapshots →" and "Back to the newest". ## The three kinds of snapshot Every snapshot page opens with an "All snapshots" link back to the list and a title naming the dataset and its day, or your label. ### AI answers The raw answers one run or upload archived. The search box ("Search keyword or response text…") and the engine picker ("All engines", or one engine) narrow the table, with a counter of matching answers against the total. The columns are Date, Engine, Keyword, Mentioned (the brands the answer named, or a dash) and Citations (how many pages it linked to). Click a row to read the answer in full: the date, the engine and the country when known, the brands under "Mentioned", the "Response" text with a note that it is AI generated and for reference only, "Citations (N)" as links you can follow, and "Fanout queries" when present. ### Visibility One row per brand in the report, your brand first and highlighted, with "Mentions", "Citations", "Impressions" and "AI Share of Voice" (a percentage), over the dates the subtitle names. AI Share of Voice is an average of the daily readings; the other three are totals. The "All models" / "Per model" toggle switches between one table that adds the engines together and one table per engine. "Compare to another snapshot" opens the comparison with this snapshot as A. ### Cited pages A read-only view of the cited pages this run or upload brought in. A bar reads "Read-only historical snapshot. Edit in Opportunities." with an "Open Opportunities" button; there are no status edits, AI actions or bulk actions here, but you can filter and choose columns. A row is an opportunity: see [What an opportunity is](https://www.sourcesignal.ai/docs/opportunities/what-an-opportunity-is.md). Under the header sits a "Briefing" card: "AI analyst briefing: read the snapshot in 30 seconds." Press "Generate briefing"; the toast "Briefing generated" confirms it. The card's top line then shows the briefing's headline. The body runs "Data maturity", "What changed", "Top movers", "Priorities", "Risks & blockers", "Route review" and "Next steps" (each only when it has something to say), then a "Rationale" disclosure. Lines under "What changed" carry a confidence chip such as "High confidence". The card offers "Export" ("Download CSV" or "Print / Save as PDF"), "Share" and "Regenerate", and shows a "Stale" chip when the data behind the briefing changed after it was written. See [Sharing, printing and briefings](https://www.sourcesignal.ai/docs/reports/sharing-printing-and-briefings.md). ## How to compare two snapshots Comparison works on visibility snapshots and needs at least two in the project. 1. Press "Compare snapshots" in the header, or "Compare to another snapshot" on a visibility snapshot. 2. Pick "Snapshot A" and "Snapshot B" by label. From the header, A starts as the oldest visibility snapshot and B as the newest. From a visibility snapshot, A is that snapshot. 3. Switch the toggle to "Per model" for one table per engine. The page opens on "All models"; if that view reads "No data for this section.", "Per model" still shows the figures. 4. Read the table. Each metric column is headed "Mentions (A → B)" and so on, and each cell shows the A value, the B value and the change, with AI Share of Voice changes in percentage points. Gains are green, drops red, your brand first. For a trend over time rather than two points, see [Visibility and share of voice](https://www.sourcesignal.ai/docs/reports/visibility-and-share-of-voice.md). ## How to upload a CSV snapshot An upload brings an Ahrefs Brand Radar export into the project as a snapshot. The dialog asks "What's in the file you're uploading?" and offers three types: | Type | What the file holds | Lands as | |---|---|---| | "Outreach opportunities" | one row per cited URL with its domain, prompt and labels | a Cited pages snapshot | | "Visibility metrics" | a Brand Radar Overview export: brands by dates by metrics | a Visibility snapshot | | "AI responses" | one row per keyword, model and date, with the columns Date, Model, Keyword and Response; a column ending in "Mentioned" is read as a brand-mention flag | an AI answers snapshot | 1. Press "Upload snapshot" and pick the type. A strip under the title names your file and the type, with "Change data type". 2. Add the file. - "Outreach opportunities": drop a CSV on the drop zone or click to pick one, give it a "Period label" (such as "Baseline" or "Month 1"), then press "Continue". - "Visibility metrics": drop the file or click to browse, and the preview opens. A file of the wrong shape is refused with "That doesn't look like a Brand Radar Overview export" and the reason. - "AI responses": choose the file, then press "Continue". 3. Check the preview. For "Outreach opportunities", map "Domain (required)" and "URL (required)", optionally "Page title", "Prompt text" and the headers to keep as "Imported labels", press "Validate", review "Will import", "Skipped" and "Duplicates collapsed", then press "Confirm import". For the other two types, review the "Date range", the row counts and any skipped rows, give the snapshot a "Period label", then press "Import". AI responses start with "Responses to " in the field; for visibility you type one, and the hint suggests "Apr 2026", "Week 17" or the date range. 4. Keep the window open. The dialog reads "Importing N … into