The Mentions feed

The Mentions page is where you read the actual evidence behind your AIS Score — every prompt where your brand was named, which platform said it, how it was framed, and how that’s shifted over time. It’s organised as three tabs: Mentions, the browsable list; Searches, the web searches each AI engine ran while answering; and Changes, what’s moved recently.

​

The Mentions tab

Mentions are grouped by prompt, not listed one row per scan. Each prompt row shows the platforms that mentioned you, an average sentiment score, a mention count with average position, and the date you were most recently mentioned for that prompt. Click a row to expand it into a per-country breakdown, or click the prompt text itself to open its full prompt detail page. Above the table, four summary cards give you the headline numbers for the last 30 days — mentions, citations, average position, and average sentiment — each compared against the prior 30-day window so you can see whether things are trending up or down. To find a particular prompt, type in the search box above the list. It matches the prompt’s wording or any of its tags, the same way search works on the Prompts page. Search only narrows the list. The summary cards stay the same, and keep following the Branded / Non-branded and tag filters.

​

Reading a mention

Expand a prompt group and then a specific scan to open the response drawer — the full text of what the AI actually said, with your brand’s mention highlighted in context. This is the fastest way to understand why a sentiment score or position landed where it did, rather than taking the number on trust. Where the engine reports them, the drawer also lists the web searches it ran before writing that answer — see The Searches tab for what these are and which platforms expose them. Use the country and platform tabs at the top of the drawer to switch between answers. When a platform has answered the prompt more than once, a Scan history picker sits under the tabs and shows how many scans there are. Each scan in the list shows when it ran and whether it mentioned you, with your position if it did, so you can open an older answer and see what changed. While you’re reading an older scan, the drawer says so, and Back to latest takes you back to the newest one.

​

Per-country drill-down

Every prompt can run in more than one market — see Countries & markets for how country targeting works. Expanding a prompt row breaks its results down by country, then by platform within each country, so you can see at a glance whether you’re mentioned consistently everywhere the prompt runs or only in some markets. This matters because a prompt scanned across three countries with a single average sentiment score can hide a real split — strong in one market, absent in another.

​

The Branded / Non-branded filter

The same three-way filter used on Prompts — All, Branded, Non-branded — is available here too, so you can look at your Mentions feed as a whole or restrict it to one slice. Switching the filter recomputes the summary cards and the recency table together, so the numbers you see always describe exactly the slice you’ve selected. See Branded vs non-branded prompts for what the split means and why it matters.

​

Filtering by tag

Tags you add to prompts in the prompt library show up on this page too. Each prompt row shows its tags. The Tags button next to the Branded / Non-branded filter narrows the whole page to prompts that carry any of the tags you pick. You can also click a tag on a row to filter by it straight away. Like the Branded / Non-branded filter, the tag filter recalculates the summary cards and also applies to the Searches and Changes tabs, so every number on the page describes the same set of prompts. The selected tags are saved in the page’s web address, so a filtered view survives a reload and you can share it. The two filters work together. To see how your non-branded pricing prompts are doing, choose Non-branded and the pricing tag. The tags widen: a prompt only needs one of the tags you’ve selected. The Branded / Non-branded filter then narrows that set.

​

The Searches tab

Before answering a prompt, most AI engines quietly run their own web searches — often several per answer — and what they chose to search for is a signal you can’t get from the answer alone. The Searches tab shows those actual queries, captured from your scans over the last 60 days: not what the AI said about you, but what it went looking for before it said it. Each row is one prompt, showing the engines that answered it — each with its own capture state — how many distinct searches were captured, and when a search was last seen. Up to two of the real queries appear under the prompt text, and a recurring badge marks a query seen on two or more separate scan dates — a strong hint the engine treats it as core to answering that prompt. Click any row to open the response drawer, which lists every search behind that specific answer. You rank for is the column to watch once you’ve connected Google Search Console. For each prompt, it shows how many of the captured searches your site already earns impressions for in ordinary Google search — “3 of 5” means the engines searched five things and Google already shows you for three of them. The gap is what matters: it counts queries an AI engine actively uses to answer a question you care about, where your site currently doesn’t appear. The check runs against your synced Search Console data for the countries this brand tracks, over the last 90 days; without a connected, mapped property with recent data, the column shows a dash rather than a misleading zero.

​

The Changes tab

The Changes tab answers a different question: not “what’s my current mention picture” but “what’s moved in the last 30 days.” It surfaces two feeds: What changed — a chronological list of specific, named transitions: a mention gained on a platform where you weren’t previously appearing, a mention lost, a position that improved or dropped, a citation gained or lost, or sentiment moving up or down. Each entry names the prompt, the platform, and the shift, so you’re reading concrete events rather than inferring them from a trend line. Slow drift — signals that build gradually rather than flip in a single scan: a prompt whose visibility has been drifting up or down over the last week compared with the two weeks before it. Drift catches the kind of change that’s easy to miss scan-to-scan but adds up — a platform that’s been mentioning you slightly less often each week, for instance. The data behind both feeds is refreshed within a few minutes of a scan completing, so opening the tab — or coming back to it — after a morning scan shows that morning’s transitions; you don’t need to wait until the next day to see what moved.

​

Related pages