Gap diagnosis

The gap diagnosis lists every prompt where visibility is weak and explains the shape of the problem. Well-cited prompts do not appear here.

The tab has two parts. On the left is the list of gap prompts; on the right is the citability audit of the prompt you pick. On a phone the list comes first, and tapping a prompt opens its audit as a sheet from the bottom of the screen.

The three gap classes

1. Absent: no AI answer for this prompt cited or mentioned you at all
2. Mentioned, not cited: answers name your brand but never link your site
3. Inconsistently cited: the prompt ran at least three times and you were cited in a third of the runs or fewer

Each class points to a different fix. Absent prompts usually need content that does not exist yet, or better retrieval. Mentioned-not-cited prompts usually need citability work: pages that are worth linking. Inconsistently cited prompts need anchor content that makes the citation stick.

Filtering the list

Filter chips above the list narrow it to one gap class, each with its count:

• All
• Absent
• Mentioned, not cited
• Inconsistently cited
• Not visible anywhere

Not visible anywhere shows the prompts that ran at least twice in the period without a single appearance of your brand on any platform: no citation and no mention. These are absent prompts with enough runs to rule out a fluke, which makes them the clearest content gaps you have.

The list is sorted by runs, most first, and shows the first 25 prompts; Show more adds 25 at a time. When the brand has search data (see below), a Sort switch above the list offers Runs and Search demand. Search demand puts the prompts with the most search impressions behind them first; prompts with the same impressions keep the order by runs. The filter, the sort, the number of rows and the selected prompt are part of the page address, so a link to a filtered list or to one audit opens exactly that view.

What each row shows

Each row shows the prompt, its gap class, its topic and number of runs, its search demand when the brand has search data, and the audit status: Audited, Audit running or Not audited. A circular arrow next to Audited means the inputs changed since that audit, for example a new winning page or a newly pinned target page; you can re-run it from the audit.

For Claude and ChatGPT runs a grey note on the row shows, per platform, in how many of the runs where the AI searched it fetched a page on your domain, and in how many of those it then cited you. Never fetched is a retrieval problem: the page is not found. Fetched, not cited means the page was read and rejected, so the fix is citability, not retrieval. The two can differ per platform for the same prompt. This note is the only place where SEOAI shows what an AI actually fetched; the audit result classes further down are based on a Google check instead.

Search demand per prompt

When a Google Search Console property or a Bing Webmaster Tools site is mapped to the brand and synced, every row also shows the search demand behind the prompt, for example "1,240 impressions · pos 4.2". That is the sum of the impressions of all search queries the prompt covers, and their average position, weighted by impressions. A prompt covers a query by the same rule as on the Search tab: the meaningful words of the query are found in the prompt or in the sub-queries the AI researched for it (fan-out), with place names and filler words ignored. Branded queries and place-name-only queries never count. No search demand means the prompt covers none of the source's queries. The position is left out when the source reports none for the covered queries.

The numbers come from one source, never added up: Google Search Console when it is mapped and synced, otherwise Bing Webmaster Tools. Copilot exports are not included, because they report citations, not impressions. The line above the list names the source and its window, for example "Search demand: Google Search Console, 28 days ending Oct 2, 2026". The audit next to the list repeats the selected prompt's demand in its header line, for example "1,240 impressions, pos 4.2 on Google".

Search demand is shown on Pro and Agency. Without a mapped and synced property or site, the demand line, the window line and the Sort switch are not shown; connect one under Settings, Integrations.

From class to audit

Select a prompt to see its citability audit next to the list. Until you pick one, the first prompt in the list that already has an audit is shown. A prompt without an audit shows what the audit does and an Audit this prompt button that runs it against the competitor that wins the prompt; no button is offered when no other page is cited for the prompt yet, or when the winner is a video. A running audit shows its state and the view updates on its own when it is done.

The audit puts your page next to the winning page and returns the best-matching passage on each side, the concepts your page misses and a suggested answer sentence, which turn the gap class into a task list. The numbered fixes and the pages the AI relies on are folded away under the summary; open them when you are ready to act. Treat the result as a hypothesis: the next runs show whether a change actually moves the citation.

The audit result classes

The audit also checks your domain in Google search, as a stand-in for whether AI search can find your content at all:

1. No content: a site search for the topic finds no page on your domain
2. Not retrieved: you have pages on the topic, but none ranks in Google's top 30 for the prompt, so AI search is unlikely to pick them up; typically a retrieval or indexing problem
3. Retrieved, not cited: one of your pages ranks in Google's top 30 for the prompt, yet the AI answers cite others; typically a content or authority gap

These classes are an estimate from Google rankings, not a record of what the AI read. For Claude and ChatGPT, the grey note on the gap row shows what the AI actually fetched.

On Free and Starter

Free and Starter plans see the Not visible anywhere prompts as a plain list on this tab. Gap classes and audits are part of Pro and Agency.

Local presence and footprint

Two checks that are not about a single page live in Citations and sources: the local presence gap, for prompts where Google shows a local pack without you, and the external footprint, which compares how known your domain is with the competitors'.