Updated
Updated What changed on
Fact-checked The UK chatbot shares, zero-click rates, Ahrefs AI Overview citation split, Claude web search details, AIClicks address, Goodie address and GEO paper details were re-checked against the 14 October 2026 source checks. Nothing else on the page was re-checked.
- Corrected the UK AI chatbot shares: StatCounter measures referrals, not queries, and in September 2026 had ChatGPT at 80.02%, Gemini 10.14%, Copilot 5.61%, Perplexity 2.7% and Claude 1.51%.
- Replaced the 58.5% zero-click figure with SparkToro's June 2026 numbers: 69.5% in the UK and 68% in the US.
- Replaced the claim that AI Overview citations follow the top three results with Ahrefs' March 2026 split: 37.9% in the first 10 results blocks, 31.2% at positions 11 to 100 and 31.0% beyond the top 100.
- Updated the Claude row: removed the retiring Sonnet 4.5 and the old web search toggle advice, and added that Anthropic lists Brave Search as Claude's web search provider.
- Fixed the AIClicks link, which now points to aiclicks.io.
- Corrected the GEO paper's authors and method: Princeton, IIT Delhi and other researchers, published at KDD 2024, measuring visibility within five supplied sources rather than citation lift.
- Fixed the Goodie link, which now points to higoodie.com. The old goodie.ai address is a parked domain listed for sale.
What is AI search visibility, and how do you actually measure it?
The 30-second answer
AI search visibility is whether your brand shows up inside the answer an AI engine composes, not whether your URL ranks somewhere on a results page. Three numbers carry the weight: how often the engine cites you (citation frequency), what percentage of total brand citations are yours versus named rivals (share-of-voice), and the count of distinct priority prompts where you appear at all (coverage). Sample weekly or fortnightly across the five engines that matter for UK B2B in 2026: ChatGPT, Perplexity, Claude, Gemini and Google AI Overviews. Anything more frequent than weekly produces noise because the outputs drift session-to-session.
How visibility is measured, and how much the measurements disagree
of URLs cited in Google AI Overviews appeared in the first 10 results blocks for that query. Ahrefs counts blocks, not organic results, so this is not the same as ranking in the top 10.
How it was measured: 863,000 keyword SERPs and 4 million AI Overview URLs. Ahrefs measures the top 10 as blocks, counting ads, featured snippets and video packs as separate blocks.
was Ahrefs' July 2025 figure. It counted only the top 3 citations per AI Overview, and Ahrefs has since changed its parsing, so the two years are not like for like.
How it was measured: 1.9 million citations from 1 million AI Overviews, analysing only the top 3 most visible citations in each one. Ahrefs notes it improved its parsing between this study and the 2026 update. Both differences mean the fall to 37.9% cannot be read as a like for like change.
of Google searches trigger an AI Overview, depending entirely on whose panel you read.
How it was measured: Pew measured 18% on real-user browsing data. Ahrefs measured 48% of its keyword set in March 2026. Similarweb reported 43%, up from 15% a year earlier, via TechCrunch on 27 July 2026. The panels sample different query mixes, so the spread is a methodology artefact. Anyone quoting a single figure as the number is overstating what is known.
Sources disagree on this one. The range is the honest answer.
Why this is not the same problem as ranking
SEO measured a position on a list of ten links. AI engines do not return ranked lists, they synthesise a paragraph or two and either cite or paraphrase the sources they pulled from. Two consequences flow from that, and both change the metric you need to care about.
First, you can be quoted without being clicked. The user gets the answer in the panel, no visit logged, no GA4 session, but the brand recognition still happens. Second, you can be skipped entirely even though you rank position one in classic Google. The AI model decided your page did not answer the question well enough to lift a sentence from, or it preferred a source it trusted more. Rank stopped being a reliable proxy for being seen the moment AI engines started writing the first impression.
The shape of the shift in numbers
The honest read on these: the user shift is real, the citation pie is small, and almost nobody is measuring yet. That window is the cheapest it will ever be to set a baseline.
The three metrics that matter
Vendor dashboards will offer you 15 to 20 metrics. Three of them carry the weight, the rest are sub-views of those three.
| Metric | What it measures | Cadence |
|---|---|---|
| Citation frequency | Percentage of priority queries where your brand is cited (across the chosen engines) | Weekly |
| Citation share-of-voice | Your citations divided by all named-brand citations across the same queries | Fortnightly |
| Query coverage | Number of distinct priority queries that cite you at least once | Monthly |
| Citation depth | Whether you are cited as the primary source, secondary, or just mentioned | Monthly |
| Cross-engine consistency | How many engines cite you for the same query (1 of 4 vs 4 of 4) | Monthly |
| Sentiment of citation | Whether you are cited positively, neutrally or negatively (recommended vs warned-against) | Quarterly |
Citations and mentions are not the same thing
Worth pulling apart. A citation is a linked source the user can click. A mention is a name-drop with no link. Both move different parts of the funnel and the engines behave very differently:
| Engine | Citation behaviour in 2026 | What that means for you |
|---|---|---|
 ChatGPT | Mentions brands often, links to them less often. High mention-to-citation ratio. | Track brand mentions, not just linked citations. Referral traffic understates influence here. Use the how to rank in ChatGPT checklist when this engine is the main gap. |
 Perplexity | Balanced. When it mentions a brand it usually links it. | Citation share-of-voice is a reliable proxy for influence. Referral traffic actually shows up. |
 AI Overviews | Citation-heavy, brand-domain skewed (59.8% of citations land on the brand's own domain). Appears on around 33% of queries. | Strong organic SEO foundations still help, but a top ranking is no guarantee. Ahrefs' March 2026 study found 37.9% of AI Overview citations in the first 10 results blocks, 31.2% at positions 11 to 100, and 31.0% beyond the top 100. |
 Claude | Web search cites consistently: Anthropic says every web search response includes citations. Without web search, no citations at all. | Anthropic lists Brave Search as Claude's web search provider, so Brave coverage and Claude-SearchBot access both matter. |
 Gemini | Cites visibly in the response footer. Sources can be unstable between sessions. | Run each query 3 times and take the union of sources to smooth the variance. |
Engine behaviour summary draws from the Otterly 2026 Citations Report (1M+ citations across ChatGPT, Perplexity and Google AI Overviews).
What actually moves the needle
The most useful published correlation set on this question is Onely's 2026 study of seven AI-visibility factors, which ranked the signals by measured correlation with citation rates across major engines. The order is consistent with what I see in client baselines:
| Signal | What it actually means | Correlation |
|---|---|---|
| Brand mentions across the web | How often your brand name appears in third-party content (any context, linked or unlinked) | 0.664 |
| Branded anchor text | Inbound links that use your brand name as the anchor (not generic anchors) | 0.527 |
| Branded search volume | People searching your brand name in classic search | 0.392 |
| Domain authority | Aggregate link-equity score for your domain | 0.326 |
| Backlinks | Raw count of inbound links | 0.218 |
The pattern is hard to miss: off-site brand presence beats on-site SEO mechanics for AI visibility. The work that earns mentions in third-party publications (digital PR, podcasts, expert quotes, original data) outranks the work that earns links. Both still help, but PR-style activity now sits ahead of link-building for this specific outcome.
The vanity metrics to avoid
What not to track: total AI mentions, raw count of AI bot crawls, "AI visibility score" composite numbers that vendors invent to look comparable. These move with noise and rarely correlate with the citation outcomes you actually want.
Specifically:
- Total AI mentions. Includes context retrieval ChatGPT does without crediting you. Inflates the number, hides the trend.
- AI bot crawl volume. A bot crawling does not mean a citation. Bots crawl widely; citations are the actual signal you want.
- Composite "AI visibility score" indexes vendors publish to make their dashboards look like SEMrush. Without methodology transparency, comparing scores across vendors is meaningless.
- Referral traffic alone. ChatGPT mentions you 4x more than it links you (Otterly data). Tracking only clicks misses most of the brand-recognition impact.
What a baseline reading actually takes
For most brands, a credible baseline is:
- 10-20 priority queries that match real buyer intent (not branded queries, those will always cite you)
- 3-5 named competitors in the citation universe so share-of-voice is meaningful
- 4 engines minimum, plus AI Overviews. StatCounter's UK AI chatbot referral shares for September 2026 were ChatGPT 80.02%, Gemini 10.14%, Copilot 5.61%, Perplexity 2.7% and Claude 1.51%. The core set:
ChatGPT
Perplexity
Claude
Gemini
AI Overviews
- Each query run 3 times per engine to smooth output variance
- Total: ~225 query-runs to build a credible baseline (15 queries x 5 engines x 3 runs)
The free AI Search Visibility tool on this site runs a scaled-down version for one-off checks. For ongoing tracking at this volume, a paid tracker (Profound, AIClicks, Otterly, Goodie) becomes practical because the volume is too high for manual sampling at weekly cadence.
Where the methodology comes from
The three-metric framework draws on published work worth reading direct: Search Engine Land's "Measuring GEO", iPullRank's AI Search Manual on measurement, the original GEO paper by Princeton, IIT Delhi and other researchers (Aggarwal et al., November 2023, KDD 2024), which measured how visible a page was within a fixed set of five sources, tested on about 1,000 queries, the Otterly 2026 Citations Report for per-engine behaviour, and Onely's correlation work for what actually moves citations. The visibility-as-citation framing (rather than visibility-as-traffic) is also explicitly endorsed by Google's May 2025 AI search guidance.
More on measurement.
How is AI search visibility different from SEO rank?
Rank is a position on a results page. AI search visibility is whether your brand appears inside the synthesised answer. AI engines do not return ranked lists, they return composed paragraphs. The measurement is fundamentally different: citation frequency, share-of-voice across named brands, and how many of your priority prompts cite you instead of position one through ten.
What is a good AI visibility score?
No industry benchmark exists in 2026 because the discipline is too young and vendor scores are not comparable across tools. Practical reading: if fewer than 20% of your tracked priority prompts cite you across two or more major engines, you have a clear gap. If you appear in over 60% across three engines, you are doing well by 2026 standards.
Which AI engines should I measure across?
For UK B2B audiences in 2026, the core set is ChatGPT, Perplexity, Claude and Gemini, plus Google AI Overviews which sits inside Google Search rather than as a standalone engine. StatCounter put ChatGPT at about 80% of UK AI chatbot referrals in September 2026, with Gemini second at about 10%, Copilot 5.61%, Perplexity 2.7% and Claude 1.51%. Copilot sits ahead of Perplexity and Claude on that measure, so it is worth adding. AI Overviews currently appears for around 33% of queries (Otterly 2026 Citations Report), so combining the standalone engines with AIO gives the fullest picture.
Should I track AI visibility daily?
No. Daily creates noise because AI outputs vary per session, per prompt phrasing and per user context. Weekly or fortnightly sampling is the right cadence. Trends across 8 to 12 weeks matter far more than single readings.
Do citations and mentions count the same?
No, and you should track them separately. A citation is a linked reference that can drive measurable referral traffic. A mention is a name-drop with no link. ChatGPT mentions brands far more often than it links to them; Perplexity has a more balanced citation-to-mention ratio; AI Overviews skews citation-heavy (around 59.8% of AIO citations go to brand domains per Otterly). Both matter, both move different parts of the funnel.
Which tools measure AI search visibility?
Free baseline: the AI Search Visibility tool on this site. Paid trackers: Profound, AIClicks, Otterly.AI, Goodie, plus visibility modules now bolted onto SE Ranking, Semrush and Ahrefs. The honest comparison is on the rank-trackers listicle on this site.
Is AI visibility correlated with revenue?
Indirectly. AI visibility correlates with influenced pipeline and brand recognition more than direct attributable revenue. Buyers reference AI engines during research and arrive at a brand they already trust as a result. The path is harder to attribute than a direct click but the influence is real, growing, and most clearly seen in branded search lift after sustained citation gains.
Jason Burns
Independent UK SEO, GEO and AI consultant. In search since 2007. Portfolio includes 3M, BlackRock, Unilever and E.ON. Owner of SEO Moves Ltd since 2014.
More about Jason →Want a proper baseline reading?
Send your 10 priority queries and three competitor names. I run them across five engines, three times each, and send back the citation frequency, share-of-voice and query coverage numbers as a one-page report with the per-engine breakdown shown above.
Request the baseline