The headline figures from the latest AI chatbot market-share report are striking: ChatGPT still leads with 54.7% of worldwide web visits across the seven largest assistants, Google's Gemini has surged to 27.4%, and Anthropic's Claude — third at 8.2% — grew its web traffic about 306% in a single quarter. The numbers, compiled by Momentic from Similarweb data, are real and widely requoted.
They are also measured with a yardstick that the assistants themselves are quietly snapping in half.
The metric is "web-visit share" — each assistant's slice of traffic to its consumer website. That is an inheritance from the search era, when popularity meant visits to a page. But generative assistants are built to do the opposite of sending you to a page: you ask, they answer, and the click-out increasingly never happens. Ranking AI assistants by website visits in 2026 is a little like ranking television networks by how many people walk to the studio. The ranking is not meaningless, but it is measuring the wrong building.
What the table actually captures
Start with the parts that hold up. ChatGPT's web-visit share has slid from roughly 76.5% in February 2025 to 54.7% now, a loss of more than 20 points as the market fractured from a near-monopoly into a real contest. Gemini's rise to 27.4% is the clearest countertrend, and Claude's 306% jump — from about 203 million web visits in January 2026 to 824 million in April — is the fastest growth of any major assistant, off a smaller base. In the United States, Claude's share climbs to 12.5%, ahead of its global average. Behind the leaders, DeepSeek (about 4.1%) and Grok (about 2.8%) are larger than most casual rankings admit.
As a directional signal, that is genuinely useful: the trend lines show where momentum and distribution are moving. The problem is treating the percentages as a scoreboard of total usage, because website traffic now captures a shrinking and uneven fraction of how people actually use these tools.
The zero-click shift the numbers can't see
The deeper context is a change in search behavior that has been building for two years. Search is being internalized: instead of returning a list of links to click, the system returns a synthesized answer, and the visit to a source site simply does not occur. The data on Google's own results makes the scale concrete. By 2026, about 65% of Google searches end without any click to an outside site, per SparkToro and Datos clickstream data. On queries that trigger an AI Overview, the zero-click rate rises to roughly 83%, and in Google's conversational AI Mode it reaches about 93%, according to Semrush. The mechanism behind that, introduced at Google I/O in 2025, is "query fan-out": one question is automatically broken into many sub-searches and recombined into a single answer, producing few or no outbound clicks.
If that is what is happening to web traffic on the world's largest search engine, the same logic applies to a panel that ranks AI assistants by visits to their websites. The more successful an assistant is at answering you in place, the less it generates the very web traffic the ranking counts. The metric is, in a real sense, structurally biased against the products it is trying to measure.
Why the assistants aren't even comparable on this axis
There is a second, subtler problem: the seven assistants do not behave the same way, so their "web visits" are not measuring the same thing. They differ in how much they keep you on their own surface versus push you out to sources, in how often they cite at all, and — critically — in where the usage lives. A large share of OpenAI's and Anthropic's real token volume runs through enterprise APIs that never touch a consumer website. Microsoft's Copilot usage happens mostly inside Windows, Edge, and Microsoft 365, not on a standalone site. Mobile apps, where all the big assistants have substantial audiences, are invisible to a web-traffic panel. Each of those is a different leak in the same bucket, and they leak at different rates for different companies. Anthropic's enterprise revenue, for instance, has reportedly grown faster than its consumer web numbers imply — which is exactly what you would expect if web visits undercount an assistant with a heavy API footprint.
This is why a growing body of marketing practice has stopped tracking page rank and started tracking "AI Share of Voice" — how often a brand or answer appears, how clearly, with what attribution, and in what position inside the AI's response. The same shift applies to the assistants themselves: the meaningful question is not how many people visited claude.ai, but how often Claude is the surface that actually answered, wherever that happened.
What actually moves share: distribution, not visits
If web visits are a fading proxy, what is the real contest? Distribution. Gemini's climb to 27.4% tracks Google's owned channels almost exactly — Search, Android, Workspace — and, as of Apple's WWDC 2026 keynote, a custom Gemini model now powers a rebuilt Siri across roughly 2.2 billion Apple devices, with users able to choose Gemini, ChatGPT, or Claude as the brain behind Apple Intelligence. None of that shows up cleanly as "web visits," yet it is the single biggest determinant of which assistant billions of people will use by default. Claude's surge, similarly, is less a website story than a spillover from the developer and enterprise mindshare it built through Claude Code and its high-end models — usage that mostly lives in tools and APIs, not on a web page.
The companies understand this, which is why they spend on defaults, operating-system integration, and being the cited source inside other products rather than chasing site traffic. The web-visit table is the scoreboard the press can see; the real game is being the answer.
Bottom line
ChatGPT's 54.7% lead, Gemini's leap to 27.4%, and Claude's 306% quarter are real, and as trend lines they tell you something true: a near-monopoly has become a three-way race with two credible challengers behind it. But the metric is a relic. These assistants are zero-click by design, they are not comparable on website traffic, and the contest that decides the next phase is distribution and share-of-answer, not visits. Read the table as a weather vane, not a scoreboard.
Frequently Asked Questions
Which AI chatbot has the most market share in June 2026? By worldwide web visits, ChatGPT leads at 54.7%, followed by Gemini at 27.4% and Claude at 8.2%, per Momentic's analysis of Similarweb data. But web-visit share excludes apps, APIs, and embedded usage, so it understates total use.
Why is web-visit share a flawed measure? Generative assistants answer in place rather than sending you to a website, so the more useful they are, the fewer web visits they generate. By 2026 about 65% of Google searches already end with no click, and the figure is far higher on AI surfaces.
What is driving Gemini's and Claude's growth? Distribution and mindshare. Gemini rides Google's Search, Android, and Workspace channels and now Apple's Siri; Claude's growth spills over from developer and enterprise adoption that lives mostly in tools and APIs.
What is AI Share of Voice? An emerging metric that tracks how often, how clearly, and in what position a brand or answer appears inside AI responses, replacing the old page-rank-and-clicks model.
Sources: Momentic/Similarweb June 2026 report; SparkToro/Datos and Semrush zero-click data; Princeton GEO research; TechTimes WWDC 2026 coverage.
ⓒ 2026 TECHTIMES.com All rights reserved. Do not reproduce without permission.
