Recent Snapshots
One row per snapshot, per domain:
AI Volume is refreshed on every snapshot, so a prompt’s figure follows how much it’s asked about now.
Because snapshots accumulate weekly by default, this table is the trend for the prompt. Read down the Mentions and Citation Rate columns — a prompt where both are climbing is one you’re winning.
Model is worth watching. Citation behaviour shifts when a platform changes model, so a step change in your numbers that coincides with a new model name is a platform change, not something you did. Checking this column first will save you investigating the wrong thing.
Quality flags responses that were thin or off-topic. A low-quality response with zero mentions isn’t evidence of a problem — the answer itself was poor.
Cited sources
A second table lists the sources the AI drew on:
This is the most directly actionable table in the GEO section. It names the sources the model treats as authoritative for your prompt.
Look at what’s there. If the same three industry publications or comparison sites recur across prompts, being covered by them is worth more than another blog post on your own site — you’re trying to influence what the model reads, not what it crawls from you.
If your own domain is absent while competitors appear, compare their cited page against yours. Usually the difference is directness: their page answers the question in the first paragraph.
LLM Response Viewer
The full text of the AI response, with its Cited Sources listed beneath. Read a few of these rather than only the metrics. The response shows how your brand is framed — recommended, mentioned in passing, or listed as an alternative to someone else. A mention that positions you as the second-choice option counts the same in the metrics as one that recommends you outright, and they’re not the same result. It also shows how the question was interpreted. AI answers frequently address a broader question than the prompt as written, which tells you what content would actually be relevant.Check now
Check this prompt now asks every AI engine the prompt again, outside its weekly schedule, and records a new snapshot. Click Check Now. A check outside the schedule spends a research lookup, so it asks first: the confirm says what it does, that it uses 1 research lookup, and how many you have left this month. Click Check Now in the confirm to queue it. The page updates when the new snapshot is recorded.Export
Recent Snapshots and Cited sources each have a ⋮ menu in their toolbar, next to the search box, or in the card’s title row when the table has no search box. Choose Export as CSV (opens in Excel and Google Sheets) or Export as JSON. The file holds every row matching the search, not just the page on screen, with the table’s own columns, up to 10,000 rows. Exports read stored data, so they’re free.Using it
- Read the Recent Snapshots trend — is the prompt improving?
- Check Model before attributing any change to your own work.
- Read the cited sources — who owns this prompt?
- Read one or two full responses — how are you framed, and how is the question being read?
- Act on the sources, not just your own pages.
Related
- GEO overview — managing prompts and the schedule
- AI competitors — the same picture from the competitor side
- Citation gap — a systematic comparison across keywords and competitors