In short
AI search market share depends on what you count. Similarweb estimates that ChatGPT received about 53% of worldwide generative AI website visits in May 2026, versus roughly 27-28% for Gemini and nearly 9% for Claude. Those are web-traffic shares, not shares of all searches or revenue. Perplexity ranked eighth in Similarweb's broader chatbot-and-tools website category in October. The sources below do not establish a comparable worldwide AI search revenue total.
How to interpret conversational AI market share in 2026
A market share needs a denominator. Visits to an assistant, people using it and clicks sent to other websites describe different activities. A person can use several assistants, and a website traffic series may exclude mobile apps. Adding weekly and monthly audiences does not produce a worldwide user total.
In its AI Search Stats 2026 report, published July 29, 2026, Similarweb gives these approximate shares for May 2026. Keep the period and measurement attached when quoting them.
| Platform | Share of worldwide generative AI website visits | Observation period |
|---|---|---|
| ChatGPT | About 53% | May 2026 |
| Google Gemini | Roughly 27-28% | May 2026 |
| Claude | Nearly 9% | May 2026 |
These are estimates reported in the source text, not precision measurements we have independently reproduced. They are also a partial table: we have not invented a Perplexity value or assigned the remainder to a particular competitor. Google's Gemini figure is not Google Search's share and does not measure exposure to AI Overviews.
Similarweb describes ChatGPT's share falling from about 76% in June 2025 to about 53% in May 2026 while its absolute visit count stayed roughly flat. That distinction matters. A service can lose share because the category grows around it, not because its own traffic collapses.
Perplexity AI market share in 2026: what can we verify?
The verified sources used here do not give a Perplexity percentage on the same May 2026 basis as the table above. Similarweb's October 2026 AI Chatbots and Tools ranking, updated November 1, places perplexity.ai eighth by website traffic. That is a rank in a broader category, not an AI search market-share percentage.
Why not turn that position into a share? Because the ranking does not supply the visit counts and denominator needed. Nor does an assistant's audience reveal how often it links to external websites. For a publisher or brand, the useful follow-up is narrower: does Perplexity cite your pages on relevant questions, do those links work, and do visitors convert?
AI search market size in 2026: usage is not revenue
Similarweb's July report says average monthly visits across generative AI platforms worldwide reached 9.5 billion in the June 2025 to May 2026 period, with 70% year-over-year growth. This describes the broader generative AI category. It is not a count of search queries, unique people or dollars spent on AI search.
The sources cited here do not establish a worldwide 2026 revenue estimate for AI search, or for the narrower AI search optimization market. We therefore do not repeat a dollar total assembled from unrelated forecasts. A report counting assistant subscriptions, enterprise APIs or agency services can produce very different market sizes.
| Measure | What it describes | What to check |
|---|---|---|
| Assistant website visits | Use of the measured websites | Period, geography, category and app exclusions |
| AI referral traffic | Visits sent to third-party websites | Destination sample and attribution |
| Google AI Overviews exposure | Searches displaying a generated summary | Query sample and feature detection |
| GEO/AEO market revenue | Spending on specified services or software | Currency, base year, scope and forecast method |
Our view: a small, reproducible measurement beats an impressive total with no defined market. Use audience estimates to choose what to investigate, not to calculate your revenue opportunity by multiplying incompatible figures.
AI referral traffic: 1.13 billion visits in June 2025
The historical benchmark remains useful when labelled correctly. In its AI Referral Traffic Winners report, Similarweb estimates more than 1.13 billion AI referral visits in June 2025, up 357% from June 2024. The analysis covers the top 1,000 websites studied worldwide. On that scope, Google Search generated 191 billion referral visits.
These are not 2026 observations, and the destination sample is not the entire web. Do not compare the 1.13 billion referrals directly with the newer 9.5 billion visits to AI platforms: one measures outbound traffic to sampled sites, the other visits within a different category and period.
Conversions: measure them instead of promising a multiplier
AI referral growth does not establish a universal conversion advantage. Compare channels on your own site with the same conversion goal, period and documented attribution method. Keep purchases, enquiries and registrations separate. Report session and conversion counts alongside percentages, especially when samples are small.
A mention rate measures appearance in answers, not visitor purchase intent. If a landing page, consent setup or attribution rule changes, record it before calling a conversion difference a channel effect. A missing referrer also does not prove that a visit came from an assistant.
Google AI Overviews: a CTR comparison, not a promise
The Seer Interactive study published November 4, 2025 reported Q3 2025 average organic CTR of 0.70% when a brand was cited in an AI Overview, versus 0.52% when an Overview appeared without citing it. It covered 3,119 informational or educational queries across 42 client organizations.
The April 2026 update expands the research to 53 brands, 5.47 million tracked queries and 2.43 billion organic impressions, with actual observations through February 2026. Its all-query table reports February organic CTR of 2.36% with an AI Overview and 3.82% without one.
Those two reports use different cohorts and segmentations. The newer all-query percentages cannot be substituted for the older cited-versus-not-cited comparison. Seer also applies a static AIO label from its latest results-page snapshot retrospectively. The results describe associations, not proof that citation caused more clicks or that every site will follow the same trajectory.
Two findings from our own Observatory
Spontaneous visibility is not the overall visibility score
The July 2026 edition of our AI Visibility Observatory aggregates 89 brands and 145 complete audits. It reports average visibility of 67/100 and spontaneous visibility of 16%. The latter comes from an entity-extraction subset and measures appearance without naming the brand in the question.
Different scales, different scopes. Do not subtract one from the other. Brands chose to be audited, so this sample does not represent the whole market. The practical lesson is to test unbranded questions rather than count an assistant repeating a name supplied in the prompt.
A cited link may fail to resolve at verification
Our study of URLs that failed to resolve at verification reports 70 cases among 26,435 checked URLs, across 23 brands. Its separate scope covers 132 audits from September 2025 to July 2026. The published overall rate is 0.26%, but the counting rule does not establish whether the counts represent distinct URLs or occurrences on a consistent basis.
We retain that reservation rather than present the rate as a proven risk per citation or a measured invention rate. A 404 or soft 404 at verification does not establish that the page never existed. Our recommendation is straightforward: check the destination and supporting evidence, not just the presence of a link.
Which metrics should guide your GEO/AEO strategy?
Use platform reach to select an initial testing set. Then keep the questions, language and search mode stable. Track spontaneous mentions, linked citations, resolving destinations, referral visits and conversions separately. No single market-share chart tells you which assistant matters most for your sector.
The public pricing page describes a catalog of more than 50 models. That does not mean every audit runs every model. The Hero Grade definition describes product dimensions rather than a universal taxonomy of market measurement. Keep those audit scores separate from audience share and business outcomes.
Further reading
Terms and definitions
AI conversion rate; AI market share; ChatGPT; Gemini; Perplexity; Claude; Language model; Mention rate; GEO; Conventional search; Zero-click search; Google AI Overviews; AEO; Visibility score; AI citations; AI share of voice; Average position; Sentiment score; Thematic coverage.
Related guides
- GEO guide
- AEO audit guide
- Google AI Overviews and visibility
- Conventional search and AEO
- GEO/AEO monitoring tools
- AI visibility metrics guide
- AEO checklist
Questions and answers
Which AI dominates the search market in 2026?
Similarweb estimates that ChatGPT received about 53% of worldwide generative AI website visits in May 2026, compared with roughly 27-28% for Gemini and nearly 9% for Claude. These are approximate web-traffic shares, not shares of all searches, users or revenue. Perplexity ranked eighth in Similarweb's broader AI Chatbots and Tools website category in October 2026; that rank does not establish its percentage share.
Does AI traffic really convert better than Google?
The sources cited in this article do not establish a universal conversion multiplier for AI referrals. Compare channels on your own site using the same conversion goal, period and attribution method. Report session and conversion counts alongside rates. Mentions in AI answers do not measure visitor purchase intent.
Is traditional SEO dead?
The cited data do not support that conclusion. Similarweb's historical June 2025 benchmark estimated 191 billion Google Search referrals versus more than 1.13 billion AI referrals to the top 1,000 websites studied. This is neither a 2026 market share nor a measure of the entire web. Measure conventional search, AI answer visibility and referrals separately.
How much is the GEO market worth in 2026?
The sources verified for this article do not establish a comparable worldwide 2026 revenue estimate for GEO/AEO services or AI search optimization software. Similarweb's usage figures measure traffic, not spending. Before quoting a market-size forecast, check its market definition, base year, currency and methodology.
How do I measure my visibility in the AI market?
Keep a stable set of questions and record the models, language, market, date and search mode. Track spontaneous mentions, linked citations, whether links resolve, referral visits and conversions separately. AI Labs Audit documents a catalog of more than 50 models, but catalog size does not mean that every audit runs every model. Audit scores are not audience market shares.
Agencies, consultants and in-house teams: compare the answers generated for your own scenarios, languages and selected models. Our infrastructure, your scenarios, your expertise.