The mention to citation gapa field guide for brands in AI answers
AI engines can name your brand without citing your website, or cite your site without naming you. Large studies show the two positions rarely match. This guide explains the gap with the evidence, then gives you a ten question audit, a way to choose the right fix and a 90 day plan you can run with your own team or with us.
top mentioned brands also among top cited domains, every month
Semrush, UK and US, 2026lowest monthly overlap, ChatGPT in the UK
Semrush, October 2026US brands a month named everywhere but cited almost nowhere
Semrush, October 2026correlation between web mentions and AI Overview visibility
Ahrefs, May 2025Contents
- Two words doing a lot of workWhat a mention is, what a citation is01
- How big the gap isSeven months of UK and US data02
- The four positionsWhich list are you missing from?03
- Why each platform needs its own checkSources, brands and concentration by platform04
- Run a ten question auditA one afternoon baseline with a source log05
- Match the fix to the gapAccess, content, corrections or coverage06
- A 90 day planBaseline, one test, one decision07
- Six questions for leadershipWhat to ask before you fund the work08
Two words doing a lot of work
Most AI visibility reports blend everything into one score. The evidence says that hides the most useful finding, because AI answers create two separate kinds of presence that move independently.
Mentioned
The AI names your brand in its answer. It does not mean the engine recommended you or described you well. It means you were part of the answer text.
Cited
The AI lists a page on your domain as a source. It means your page was used to build the answer. It is a source position, not a judgement of trust or quality.
Fewer than half of the brands AI names are among the sites it cites
Semrush’s AI Visibility Index for the UK and US tracked ChatGPT, Gemini, Google AI Mode and Google AI Overviews for seven months in 2026, across 22 categories. On average, fewer than half of the top 100 mentioned brands in a category were also in the top 100 cited domains, in every month and both markets.
The gap runs in one direction. Each month 8 to 13 US domains, and 3 to 9 UK domains, were inside all four top 100 mention lists while outside all four top 500 citation lists. Semrush calls this the One Way Door. Chevrolet, Disney, Ford, Honda, Marvel, Nintendo and Toyota held it in every US month. The reverse, cited widely but rarely named, held zero or one domain.
Overlap between mention and source lists
| Platform | UK range | UK months below 50 | US range | US months below 50 |
|---|---|---|---|---|
| ChatGPT | 28.7 to 32.3 | 7 of 7 | 32.8 to 35.8 | 7 of 7 |
| Gemini | 33.4 to 36.9 | 7 of 7 | 37.7 to 39.4 | 7 of 7 |
| Google AI Mode | 34.3 to 52 | 5 of 7 | 43.4 to 54.6 | 5 of 7 |
| Google AI Overviews | 44.9 to 63.8 | 4 of 7 | 47 to 62.1 | 2 of 7 |
Domains shared out of 100. January to July 2026. Source: Semrush AI Visibility Index, UK and US, October 2026.
One Way Door vs its reverse, July depth test
July 2026. The qualifying cutoff varies; absence stays fixed outside the opposing top 500 on all four platforms. Source: Semrush AI Visibility Index, UK and US, October 2026.
Find the list you are missing from
Your next move depends on which measure lags. Place your brand in one of four positions for each platform and question set, then follow the matching check. This is a way to choose the next question, not a new score.
Named and cited
Find the pages and questions behind both positions and check that they recur across runs. These are the assets to protect and monitor first.
Named, not cited
Open the pages that are cited instead, including other sites. Is your brand named while another domain supplies the facts? Decide whether to fix your own page, its access, or your coverage on the cited sites.
Cited, not named
Read the answers that cite you. Check how your brand is described and whether the cited page makes clear who you are and what you offer.
Neither
Start with discovery. Find which brands and sources do appear, widen the question set, and check crawler access before choosing any tactic.
Adapted from the Semrush AI Visibility Index framework. A conceptual guide, not a measured split of brands.
One question, four different answers
The four platforms do not share one view of a category. Gemini names the most brands per answer but lists the fewest sources. ChatGPT lists the most sources and leans on news sites, especially in the UK. AI Overviews has the most distinctive set of brands and sources, and its source list overlaps least with ChatGPT’s. AI Mode concentrates a third or more of its citation weight on ten domains.
For a brand that means four baselines, not one. It also means a fix can work on one platform and do nothing on another, so test changes against the same platform you measured.
Distinct brands per answer, UK
Source: Semrush AI Visibility Index, UK and US, October 2026. AI Overviews uses a separate keyword based prompt pool.
Leading source type by platform, seven month mean
| Platform | UK leading source type | US leading source type |
|---|---|---|
| ChatGPT | News and information, 29.5% | News and information, 19.6% |
| Google AI Overviews | Streaming and online TV, 19.4% | Streaming and online TV, 24.9% |
| Gemini | News and information, 19.9% | Streaming and online TV, 17.2% |
| Google AI Mode | Business software and SaaS, 17.7% | Business software and SaaS, 20.1% |
Share of top 100 citation weight. Labels classify the domain, not the cited page; AI Mode’s software share includes social and community platforms such as Facebook, Instagram, Pinterest and Quora. Source: Semrush AI Visibility Index, UK and US, October 2026.
Run a ten question audit this afternoon
Pick ten questions real buyers ask, from sales calls, search data or support tickets. Run each on ChatGPT, Gemini and Google AI Mode, then search it on Google and record any AI Overview. Keep the market and language fixed and log the date, every cited URL and its domain, including when no AI Overview appears.
Sort what you find into three groups: your own domain, third party pages that mention you, and third party pages that leave you out. The third group is an investigation list, not an outreach list. Start with pages that recur.
Your own domain is cited
Check the page is current and says clearly who you are.A third party page that mentions you is cited
Check the description is accurate.A third party page without you is cited
Check relevance and editorial fit before any outreach.Ten questions give a direction, not a benchmark. Run each question at least twice, because answers change between runs.
Match the fix to the gap
"Create more content" is not a plan. Each finding from the audit points to a specific problem and the team best placed to test a response. If a key page is blocked or out of date, start there. If the answer cites another site, look at that page before assuming your own site is the only place to act.
| What the audit finds | A test to run | Who leads | In our GEO Pilot |
|---|---|---|---|
| A relevant page cannot be reached by the AI crawler or does not render its text | Fix robots.txt rules, rendering, status codes or canonicals, then check access again with logs and testing tools. | SEO and developers | Included |
| Your page lacks the specific, current facts buyers ask about | Add the missing facts, numbers, prices, dates and named sources to that page, then rerun the same questions. | Content and subject experts | Included |
| A cited third party page describes you wrongly | Ask the publisher for a factual correction through its normal editorial process. | Communications | We prepare the request |
| Competitors appear in cited third party pages and you do not | Check the publication’s audience and editorial fit before testing an outreach or PR idea. | PR and partnerships | Plan only |
| The gap depends on which domain address you track | Fix the measurement scope, for example www vs bare domain or country sites, before changing the plan. | Analytics and SEO | Included |
First three columns adapted from the Semrush AI Visibility Index. The last column shows what our fixed price GEO Pilot covers.
A baseline before a bet: the 90 day plan
Ask for one bounded decision at the end of 90 days: expand the test, change it or stop. The first month fixes the comparison. The second chooses one intervention. The third measures it on the same basis.
The work
Choose the market, competitors and question set. Build separate mention and citation baselines per platform. Record the collection conditions. Fix any access problems found.
What leadership gets
A dated comparison with competitors, the most important unanswered question and an agreed rule for judging the test.
The work
Inspect the answers and cited pages. One owner proposes one test with its cost, risks and dependencies, such as rewriting three pages or correcting two third party descriptions.
What leadership gets
A specific intervention and the reason it is worth testing, not a general content brief.
The work
Run the test. Repeat the same measurement on the same basis and note anything else that changed and could affect the result.
What leadership gets
A recommendation to expand, revise or stop, with the evidence and the uncertainty that remains.
Six questions for leadership
Where are our priority competitors ahead of us, in mentions and in citations?
Ask for both lists, per platform.
1How does that differ by market and by platform?
A UK plan should be tested in the UK, not copied from the US.
2Has our position changed on a comparable basis?
Same questions, same platforms, same depth, dated.
3Which questions and cited pages explain where to look next?
Including third party sites, not only our own pages.
4What test are we funding, who owns it and why?
One intervention, one owner, one business objective.
5What evidence, by what date, would make us expand or stop?
Agree the rule before the work starts.
6How GrowingAI applies the gap method for clients
Our 30 day GEO Pilot covers the first month of this plan for a fixed $999. We build the question set with you, record mentions and citations separately on each platform, fix crawler access and on page problems, and rewrite your priority pages with the facts buyers ask for. You get a dated before and after report with the gap per platform.
We do not promise rankings or citations, because no one controls them. What we can promise is a clean baseline, the fixes done in your site, and a clear recommendation for the next test. Teams that want to run it themselves can use the free checker and templates on this page.
Cite or share this research
You may quote any figure with a link to this page. Please also credit the original study named on each card.
GrowingAI. (2026, 18 October 2026). The mention to citation gap: a field guide for brands in AI answers. https://thegrowingai.com/research/mention-citation-gap/
Sources
- Semrush AI Visibility Index: UK and US, Semrush, October 2026 (coverage by PPC Land; full report available from Semrush)
- AI Overview brand correlation study, Ahrefs, May 2025
- GEO: Generative Engine Optimization, Aggarwal et al., KDD 2024
- AI features and your website, Google Search Central
FAQ
Questions people ask
What is the mention to citation gap?
It is the difference between how often an AI engine names your brand in its answers and how often it lists your website as a source. Semrush’s UK and US AI Visibility Index found that across 22 categories fewer than half of the top 100 mentioned brands were also in the top 100 cited domains, every month from January to July 2026.
What is the One Way Door in AI visibility?
It is the term Semrush uses for brands that sit inside the top 100 mention lists on all four platforms it tracks while their own websites sit outside all four top 500 citation lists. In the US, Chevrolet, Disney, Ford, Honda, Marvel, Nintendo and Toyota held that position in every month measured. The reverse position, cited widely but rarely named, was almost empty.
Why would AI mention my brand but cite other websites?
Usually because other pages answer the question more directly or more recently than yours, because your relevant page is hard for the crawler to reach or render, or because the engine relies on third party sources such as news, reviews, video or community sites for that topic. A short audit of the cited pages shows which of these applies.
How do I check my own mention and citation gap?
Take ten real buyer questions and run each on ChatGPT, Gemini and Google AI Mode, then search it on Google and record any AI Overview. Log whether your brand is named, whether your domain is cited, and which other domains are cited. Our free AI citation gap checker does the counting and suggests the next test.
Which AI platform has the biggest gap?
In Semrush’s 2026 data ChatGPT had the lowest overlap between mention and source lists, 28.7 to 32.3 shared domains out of 100 in the UK and 32.8 to 35.8 in the US. Google AI Overviews had the highest overlap, but its source mix differs most from ChatGPT’s, so each platform needs its own check.
How long does it take to close the gap?
There is no fixed time, and no one can guarantee a citation. A sensible plan uses 90 days: a month to set baselines and fix access, a month to choose and run one test, and a month to measure it again on the same basis and decide whether to expand, change or stop.
Want us to run the audit for you?
Send your site and five buyer questions. We measure mentions and citations per platform and show you where the gap is.
- Separate mention and citation baselines
- Access and page fixes done for you
- Fixed $999, delivered in 30 days