Snapshots

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.

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.

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.

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.

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:

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 <start> to <end>" 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 <label>…" and cannot be closed until the result appears.
5. Read the result: "Imported N opportunities." (or metric rows, or AI responses), with notes such as "N unmatched keywords: these answers were still imported, but are not linked to a prompt yet." Press "View imported snapshot" or "Close".

If the upload fails, "Import failed" names the reason and "Review & retry" returns you to your file and settings. "Could not confirm the import" means a snapshot may already exist: press "Check snapshots (new tab)" before retrying.

Import issues on an uploaded snapshot

When an upload skipped rows, its row reads "N import issues". Click it to open the "Import issues" dialog, which lists each skipped row by number and reason; "Download error report" saves the full list as a CSV.

How to delete a snapshot

1. Press the trash icon on the row.
2. Read the confirmation: the snapshot record and its stored file go, along with any analyst briefing written from it, and links to it stop working. The pages, visibility metrics or AI responses it brought in stay in the project, kept by date. This cannot be undone.
3. Press "Delete". The toast reads "Snapshot deleted".

Limits

Good to know

• Snapshots made by a data update are titled by their day; the label you type on an upload is kept as written.
• "Uploaded" on a row is the day the snapshot was saved, for imports and uploads alike. The two date filters work on that day, in UTC.
• After an upload lands, the follow-up work shows in the engine status at the foot of the rail until it finishes.
• Ask Spot, or your own agent over MCP, to "Compare two periods": one figure, such as AI answers gathered or your share of voice, over the last 7, 14 or 30 complete days against the same days before. The "Share of voice breakdown" says what moved alongside a change, by engine and by topic, never what caused it. Both are reads; nothing changes. See Spot, the in-app agent.