Visibility and share of voice

Every visibility number in sourcesignal starts from the same place: the answers AI engines give to the prompts you track. The numbers differ in what they count inside those answers, and over which window.

Use this page when a number needs explaining, to you or to a client.

Answers and checks

• An answer is one AI engine's reply to one tracked prompt, kept whole: the text, the brands it names and the pages it cites.
• A check is the report's unit: one tracked prompt asked of one AI platform that returned a usable answer. The report counts a prompt on a platform once per day. If it was answered more than once that day, the check counts as a mention when any of those answers named you.

On the Tracked prompts dataset, every visibility percentage in the report is mentions divided by checks, for the period and filters you chose. The Prompts and Competitors pages count stored answers instead, over the last 30 days. So the same brand can read a little differently on each.

Mentions

A mention is an answer whose text names your brand.

• Counts: answers that name you. For a competitor, answers that name that competitor.
• Leaves out: answers that only list your site as a source. Those are citations.
• Where you see it: the "AI visibility" cover and the "Prompts visible" tile in the report, the "You" column on the Prompts page, and the "Prompt coverage" tab in Internal analysis.

"Prompts visible" reads like "12 / 15": prompts where at least one answer named you in the period, out of the same fixed set of tracked prompts. Pick one platform in the report's filter bar to read mentions for that engine alone.

Citations

A citation is an answer that lists a page as a source. The page can be yours, a competitor's or a third party's.

• Counts: answers whose sources include the page or site.
• Leaves out: whether the answer named anyone. An answer can cite a page without naming the brand behind it.
• Where you see it:
  • "Cited pages" on the Prompts page: every page an AI answer to that prompt has cited, all time. Each one is an opportunity, and the number opens those pages in Opportunities.
  • "Cited in responses" and "Cited pages" in the "Mentions & citations" view of the Competitors page.
  • On Outcomes, "Visibility lift after pitching" and a win's "The pitch behind it", which compare your mentions and citations before and after you pitched.
  • The citation gatekeepers and source pages, covered in Citation sources.

Share of voice

Three numbers carry the word "share", and they answer different questions.

Tracked share of voice is a share of attention among the brands you compare, not a market share. The comparison set is fixed. Every brand shows the day it joined, and a removed brand stays struck through with its removal date, so older periods stay comparable. With no competitor in the set, the tile shows a dash and says why.

One answer can name several brands, so shares of answers do not add up to 100%.

On the report's Competitors screen, "Current position" is your rank among the brands you compare, by tracked share of voice. "Gap to leader" is how many points you trail the leader; when you lead, it shows your margin over second place. More in Topics and competitors in the report.

Mention stance

Stance is how AI answers framed your brand when they named you. The AI review reads each answer that names you, and a pill sums up one prompt's answers for the window.

To see it, open Reports, press "Internal analysis" in the tab row, then open "Prompts won / lost". Pick a 7d, 30d or 90d window. A prompt whose answers name you in the window can carry a pill beside its buying stage.

The pill keeps a new mention that called you the weaker option from reading as good news.

Beside it, an impact badge says how much the prompt's movement matters to you. "Prompts newly won" and "Prompts newly lost" are ordered by it. A prompt with no figure is not measured, not zero.

Position

Position is where you sit inside the answers that name you.

• On the Prompts page, the "You" column adds a phrase to your rate: "usually the lead pick", "usually one of a list" or "usually a footnote". The AI review places you in each answer that names you, and the phrase is the most common place over the last 30 days. Answers that only cite your site never set a position, so a prompt where you are never named shows none.
• In the report, the competitor ranking has room for an "Avg rank" column: a brand's average position inside the answers that name it. It is not measured yet, so the column stays hidden and the ranking is ordered by share.

Position inside an answer is not the same as "Current position", which ranks brands by share of voice.

Dashes, zeros and too few checks

• A dash means nothing comparable was measured. It is never a zero.
• A 0% appears only when answers were checked and none named you.
• "Insufficient data" means too few checks to show a number. A topic, or a prompt on the Topics screen, needs at least 5 checks in the period.
• "Not visible" is a prompt status on the Topics screen: checked at least 5 times in the period and never mentioned.
• Changes are in points: 40% to 45% is +5 pts, the difference between two rates.
• Report windows end yesterday. Today's checks show separately until the day ends.

Good to know

• Category phrases. A brand whose name is also a category phrase, such as "AEO Agency", is counted with care so the phrase alone does not inflate its share. Its row in the competitor ranking carries a "Category phrase" hint with how many mentions were left out. The Ahrefs market ranking holds such names out.
• AI assistants are never competitors. AI platforms and big-tech properties are named in answers all the time, so they never join the comparison set or count in a share.
• Your brand name matters. Without it in project settings, the Competitors page cannot show your share beside theirs.
• Removing a competitor keeps its history in the report.
• Movement is association. A rise in mentions after a placement is reported as following it, never as caused by it. See Impact and outcomes.