Key takeaways
- ChatGPT holds 53.9% worldwide web-visit share in May 2026, down from 54.5% in April and 79.0% a year earlier, even though its visit count is roughly flat year over year.
- Google Gemini is second at 27.9% worldwide share, about half of ChatGPT's web visits, and up about 450% year over year (527.7M visits in May 2025 to 2.9B in May 2026).
- Claude reached 9.2% worldwide share and 952.6M web visits in May 2026, up about 855% year over year and 228% in a single quarter, the fastest growth in the set on every horizon.
- The seven assistants drew a combined 10.40 billion web visits in May 2026, up about 49% year over year, so challengers captured the market's growth while ChatGPT stayed flat.
- In the United States, ChatGPT holds 58.3% web-visit share, Gemini 19.3%, and Claude 13.4%; Claude's US share rose more than a full point in one month.
- Grok (2.4%) and Perplexity (1.3%) both lost worldwide share over the past quarter; Copilot (1.3%) is flat on the web but up about 42% year over year.
- App usage adds 15% to 60% on top of web visits depending on the assistant, so web-visit share measures the consumer web product, not total usage.
- The flagship models behind the four leading closed assistants now sit within four points on PhD-level science questions (GPQA Diamond 88.5% to 94.3%), so capability has converged at the top while price has not: DeepSeek's open-weight model runs at roughly an eighth of GPT-5.5's output price.
Which AI chatbot has the largest market share in 2026?
ChatGPT has the largest AI chatbot web-visit share in 2026, at 53.9% of worldwide web visits among the seven largest generative AI chatbots in May 2026. Together the seven drew a combined 10.40 billion web visits that month. The market is two-tiered. ChatGPT and Google Gemini together hold about 82% of measured visits, and the other five assistants split the remaining 18%.
Source: Similarweb, total (desktop and mobile web) visits, worldwide, May 2026. Shares are each assistant's visits as a percentage of the seven assistants' combined visits and may not sum to 100% because of rounding. YoY compares May 2025.
Which apps are counted, and which are not
The set is standalone, general-purpose generative AI assistants on a separable primary web domain above roughly 50M monthly web visits. It covers the large assistants in English-language markets. It excludes assistants with no separable domain such as Meta AI inside Meta's apps, companion and character apps such as Character.AI, answer aggregators such as Poe, and major non-English assistants including Chinese models such as Doubao, Ernie, and Qwen. Adding any of these would lower every share figure, most visibly at positions 4 to 7. Read the lower positions as shares of this defined set, not of all AI chatbot usage worldwide.
What is the AI chatbot market share in the United States?
In the United States, ChatGPT holds 58.3% of AI chatbot web-visit share, ahead of Google Gemini (19.3%) and Claude (13.4%), based on May 2026 Similarweb data. The US market is more concentrated at the top than the worldwide market, and it re-weights the challengers. Gemini's US share is lower than its worldwide share because its strongest markets are in Asia and Latin America. Claude's US share is higher because its usage is the most US-centric of the leaders, and it gained more than a point of US share in the single month from April to May.
Source: Similarweb, total web visits, United States, May 2026. US quarterly growth compares February 2026; US YoY compares May 2025. Shares may not sum to 100% because of rounding. Worldwide and US growth are not directly comparable because the base and geography differ.
How has AI chatbot market share changed over time?
AI chatbot web-visit share has consolidated toward Google Gemini and Claude and away from ChatGPT over the 15 months to May 2026. ChatGPT's share of the seven assistants' combined visits fell from 76.5% in February 2025 to 53.9% in May 2026, while Gemini's rose from 5.6% to 27.9% and Claude's from 1.4% to 9.2%. The series below tracks each assistant's share at quarterly snapshots.
Source: Similarweb, total worldwide web visits; each assistant's share of the seven's combined visits at each snapshot. Columns may not sum to 100% because of rounding.
A few patterns stand out. ChatGPT held 76% to 79% through mid-2025, then slid 25 points over the following nine months as Gemini and Claude scaled. Gemini's climb is the steadiest, roughly doubling its share every two quarters since mid-2025. Claude's is the sharpest recent move, from 3.4% in February 2026 to 9.2% in May, nearly tripling in a single quarter. DeepSeek's February 2025 figure (12.1%) reflects its launch-driven traffic spike, which faded over the following months and has since stabilized near 4%.
How do ChatGPT, Gemini, and Claude compare?
ChatGPT, Gemini, and Claude are the three largest AI chatbots by web traffic, and each leads on a different dimension. ChatGPT leads on size, Gemini on sustained scale among the challengers, and Claude on growth rate and engagement quality.
Source: Similarweb, May 2026. ChatGPT growth figures sit within the data's rounding band and read as roughly flat.
Is ChatGPT losing market share?
ChatGPT is losing web-visit share while holding its visit count flat. Its worldwide web visits were about 5.6B in May 2026, within the rounding band of both May 2025 (5.5B) and November 2025 (5.8B), so its absolute traffic has plateaued. Over the same year the combined traffic of the seven grew about 49%, from 6.96B to 10.40B visits. A flat numerator against a growing denominator is why ChatGPT's share fell from 79.0% to 53.9% in twelve months. ChatGPT still serves more web visits than the other six assistants combined. The pattern is share erosion from a plateau, not a decline in usage.
Is Google Gemini overtaking ChatGPT?
Google Gemini is closing the gap but has not overtaken ChatGPT. Gemini is the second-largest AI chatbot at 27.9% worldwide web-visit share, about half of ChatGPT's traffic at 2.9B versus 5.6B visits in May 2026. Gemini grew about 450% year over year and 107% over the past six months, the strongest sustained climb among the large assistants. Combined with Google's distribution across Android, Chrome, and Search, Gemini is the main competitive threat to ChatGPT's lead, though it would need to roughly double again to draw level on web traffic. Gemini's flagship Gemini 3.1 Pro also tops the consumer field on GPQA Diamond science reasoning at 94.3%, so the climb pairs a leading model with Google's reach. One caveat applies. Some of Gemini's measured web visits are driven by Google entry points such as the Google app, Search, Android, and Workspace rather than by independent demand, which can flatter both its share and its growth against a destination users navigate to directly.
How fast is Claude growing in 2026?
Claude is the fastest-growing major AI chatbot by web visits in 2026, up about 855% year over year (99.7M visits in May 2025 to 952.6M in May 2026) and 228% in the single quarter from February to May 2026. This is the largest growth rate in the set on every time horizon, and it moves Claude from a developer-associated tool to the third-largest consumer AI chatbot by web traffic, approaching one billion monthly web visits. In the United States, Claude grew about 763% year over year and 149% in the quarter, lifting its US share to 13.4%. The surge coincides with Claude Opus 4.8, which ties GPT-5.5 at the top of SWE-bench Verified for resolving real GitHub issues (88.6%), a strength that fits Claude's developer-heavy base. Claude also posts the best engagement profile of the consumer leaders, covered below.
Which AI chatbot is growing the fastest in 2026?
Ranked by quarterly web-visit growth from February to May 2026, Claude is the fastest-growing AI chatbot, followed by DeepSeek, Gemini, and Copilot. Grok and Perplexity declined over the quarter, and ChatGPT was roughly flat.
Source: Similarweb, worldwide total web visits. MoM compares April 2026, quarterly compares February 2026, six-month compares November 2025, YoY compares May 2025.
DeepSeek shows the split most clearly: down about 1% year over year but up 58% in the quarter, because its traffic dipped to a 273.2M trough in February 2026 before rebounding to 430.4M in May. Its rebound fits its model economics: the open-weight DeepSeek V4 Pro scores within four GPQA points of GPT-5.5 at about an eighth of the price, which keeps it competitive in cost-sensitive markets. Read a single growth figure as the net change between two dates, not as a steady rate.
What is the early-July 2026 trajectory?
Through the first 27 days of June 2026, every assistant's US web visits are tracking below their May daily average, a uniform early-summer softness that leaves the standings unchanged. Projected from those partials, June US visits land near 943M for ChatGPT (about -4% versus May), 318M for Gemini (-2%), and 220M for Claude (-2%), with the smaller four apps down 6% to 10%. These figures are projections from an incomplete, US-only month, so the report's headline stays on May, the latest complete month, and the June reading is offered only as a directional signal.
How much does web traffic undercount app usage?
Web visits undercount total usage by 15% to 60% depending on the assistant, because native iOS and Android apps carry traffic that never appears in web data. The platform breakdown below counts active users by device for the United States in May 2026, which is a different metric from web visits and is not deduplicated across platforms, so the columns should be read as a directional measure of where each assistant's usage lives rather than a single user total.
Source: Similarweb, active users by platform, United States, May 2026. Platforms are not deduplicated, so totals are not additive across the user base.
Perplexity is the one assistant whose app users outnumber its web users, so its 1.3% web-visit share understates its reach more than any other app in the set. ChatGPT and Grok draw close to 40% of their US users through apps. Copilot is the most web-bound at 15%, fitting its position as a surface embedded in Windows, Edge, and Microsoft 365 instead of a standalone destination. For the ranking, web-visit share stays the most reproducible measure of the consumer web product, and the app gap it leaves out is largest for Perplexity and smallest for Copilot.
Which AI chatbots have the highest user engagement?
By May 2026 engagement metrics, Grok and Gemini show the highest per-visit engagement, and Microsoft Copilot's standalone web app the lowest. Claude shows the best engagement-to-traffic ratio among the consumer leaders, with a 26% bounce rate, the lowest in the set, and 4.5 pages per visit despite ranking third in volume.
Source: Similarweb, total worldwide web traffic, May 2026.
Grok's figures are outliers at roughly 11-minute sessions and about 13 pages per visit, likely because the grok.com surface bundles activity beyond chat and is entangled with adjacent X properties. Treat both its engagement and its visit counts as directional. Microsoft Copilot's 48% bounce rate fits a standalone web domain that is not where its value is delivered.
Where are AI chatbots used the most, by country?
The United States is the single largest country for every leading assistant: 17.6% of ChatGPT's May 2026 web visits, 11.2% of Gemini's, and 23.6% of Claude's. No leader draws a majority from any one country, so AI chatbot usage is spread across many markets.
Source: Similarweb, share of each app's own total web visits by country, May 2026.
Claude is the most US-concentrated of the three leaders. Gemini is the most internationally distributed, with strong East Asian usage. DeepSeek's base is heavily Asian, which is why its worldwide share (4.1%) far exceeds its US share (1.2%).
How do the leading assistants' models compare on capability and price?
The models behind the four largest closed assistants are within about four points of each other on PhD-level science questions, so capability has converged at the top of the market while price has not. The table pairs each leading assistant with its flagship model and two benchmarks: GPQA Diamond measures PhD-level science reasoning, and SWE-bench Verified measures resolving real GitHub issues.
Data from DemandSphere AI Frontier Model Tracker, updated June 29, 2026, used with permission. Flagship models shown; Copilot routes to OpenAI and Microsoft MAI models, and Perplexity routes queries to third-party models, so neither lists a separate base model.
Two splits stand out. On coding, GPT-5.5 (88.7%) and Claude Opus 4.8 (88.6%) lead SWE-bench, while Gemini and DeepSeek trail at 80.6%. On price, DeepSeek V4 Pro is the outlier: open weights at $1.74 input and $3.48 output per million tokens, roughly an eighth of GPT-5.5's output price, while scoring within four GPQA points of it. That cheap-and-close pressure is the supply-side counterpart to DeepSeek's flat web traffic. The benchmarks describe model capability, which is a separate question from the consumer usage the rest of this report measures.
AI chatbot market share by users vs. web visits
AI chatbot market share measured by users, app downloads, or revenue would rank the field differently from web visits, and no single public source measures those consistently. Reported monthly-active-user counts come from company announcements on different definitions and dates. The platform breakdown above shows how far app usage can diverge from web visits within a single market. Revenue share is dominated by API and enterprise contracts that no consumer-usage metric captures. This report uses web visits because it is one defined, reproducible quantity that can be re-pulled and audited. Where a question is specifically about users or revenue, read web-visit share as a proxy for consumer web demand.
How is AI chatbot market share measured?
AI chatbot market share in this report is web-visit share: each assistant's monthly sessions on its primary web domain, estimated by Similarweb, expressed as a percentage of the seven assistants' combined visits. Web visits were chosen because they are one defined, reproducible quantity. A users figure requires undisclosed deduplication assumptions, and a share with no denominator is not a measurement. The window is May 2026, the latest complete month. The trend uses quarterly snapshots from February 2025. Month-over-month compares April 2026, quarterly compares February 2026, six-month compares November 2025, and year-over-year compares May 2025. Figures are worldwide unless a country is named. Estimates are panel-modeled, carry an error band rather than decimal precision, and are revised month to month, so single percentages are best read as approximate.
Why do AI chatbot market share reports disagree?
AI chatbot market share reports disagree mainly because they measure different things without disclosing it. Four methodology gaps drive most discrepancies:
- No stated denominator. A share figure given without specifying users, sessions, visits, or revenue cannot be reconciled with any measured quantity.
- Undisclosed geographic scope. A US-only ranking inflates US-concentrated apps such as Claude and understates Asia-heavy apps such as DeepSeek; a worldwide ranking does the reverse.
- Web versus app blindness. A ranking built on web visits alone undercounts app-first assistants such as Perplexity, where app users outnumber web users.
- Stale product data. Listings that attribute outdated models to an app show a dataset that is not maintained against the live market.
The corrective is disclosure rather than a better black-box number. A ranking is only as credible as the methodology a reader can inspect.
Does web traffic capture total AI chatbot usage?
No. Web traffic understates total AI chatbot usage, and it does so unevenly across apps. Web visits measure the consumer web product, not total model usage or revenue. Four categories sit outside, or distort, web-visit data:
- Native mobile apps. Every assistant in the set has iOS and Android apps, and the platform table shows apps carrying 15% to 60% of US users depending on the assistant.
- Embedded and ambient surfaces. Most Microsoft Copilot usage occurs inside Windows, Edge, Bing, and Microsoft 365, so its 1.3% web-visit share understates its true reach. Gemini is similarly embedded across the Google app, Android, and Workspace, and entry points from those products may inflate its measured web visits.
- API and enterprise consumption. A large share of OpenAI's and Anthropic's token volume runs through APIs that generate no web visit.
- Automated traffic. Panel-and-model estimates can include non-human traffic, and AI domains are common automation targets, which adds noise to any single-month figure.
This is the most reproducible public view of the consumer market, and it is not offered as a measure of total usage.
What does AI chatbot market share mean for brands and AI search visibility?
For brands, AI chatbot market share shows where to monitor AI search visibility, which is a different question from where a brand is already cited. ChatGPT and Gemini holding about 82% of consumer web traffic means those two surfaces shape most AI-mediated brand impressions, so monitoring should concentrate there. Whether a brand appears in those answers depends on other signals: being indexed and ranking on Google, which feeds Gemini and AI Overviews, and being present on the third-party sources these systems ground on, such as review platforms, community threads, and structured listings. With the leading models now clustered within a few points on capability, the usage race turns more on distribution and grounding than on raw model quality, which is why being present in the sources an assistant cites matters more than which model is marginally best this month.
Two mechanics follow. First, generative assistants present the most-published consensus about a topic as fact, which turns search, PR, and social into a single input to one machine-generated consensus. Second, as users hand more of the consideration stage to assistants, being inside an assistant's consideration set matters more than ranking on a page a person may never open. One caveat: AI-search visibility measurement is directional rather than measurement-grade, because assistant answers are non-deterministic, no public dataset of real prompts exists, and roughly 80% of prompts are unique. Track it at the topic level as a trend, and report usage-share data like this report as the defined quantity it is, while treating share of voice inside AI answers as an estimate.
Frequently asked questions
What is the most popular AI chatbot in 2026? By web visits, ChatGPT is the most-used AI chatbot in 2026, at 53.9% of worldwide web-visit share and 5.6 billion web visits in May 2026.
Which is bigger, Gemini or Claude? Google Gemini is bigger. Gemini had 27.9% worldwide web-visit share (2.9B visits) versus Claude's 9.2% (952.6M visits) in May 2026, about three times larger by web traffic.
Is Gemini overtaking ChatGPT? Not yet. Gemini reached 27.9% worldwide web-visit share in May 2026 against ChatGPT's 53.9%, about half. Gemini's share rose from 5.6% in February 2025 while ChatGPT's fell from 76.5%.
What is ChatGPT's market share in the United States? ChatGPT held 58.3% of US AI chatbot web-visit share in May 2026, ahead of Gemini (19.3%) and Claude (13.4%).
Which AI chatbot is growing the fastest? Claude is growing the fastest, up about 855% in worldwide web visits in the year to May 2026 and 228% in the prior quarter.
How has ChatGPT's market share changed since early 2025? ChatGPT's share of the seven assistants' combined visits fell from 76.5% in February 2025 to 53.9% in May 2026, while Gemini's rose from 5.6% to 27.9% and Claude's from 1.4% to 9.2%.
Is DeepSeek popular outside its home market? DeepSeek held 4.1% of worldwide AI chatbot web visits in May 2026 but only 1.2% in the United States, which reflects a heavily Asian user base.
Is Perplexity gaining or losing share? Perplexity is losing web-visit share, down about 27% over the six months to May 2026, to 1.3% worldwide, though its app users outnumber its web users.
What is Grok's market share? Grok held 2.4% of worldwide AI chatbot web-visit share in May 2026, at 244.9M web visits, sixth among the seven leading assistants.
What is Microsoft Copilot's market share? Microsoft Copilot's standalone web app held 1.3% of worldwide AI chatbot web-visit share in May 2026, at 131.6M web visits, though most Copilot usage runs inside Windows, Edge, and Microsoft 365 and does not appear in web visits.
Which AI chatbot is declining the fastest? Perplexity is the fastest-declining assistant by web visits, down about 27% over the six months to May 2026, and Grok fell about 18% over the prior quarter.
How many people use ChatGPT? ChatGPT drew about 255.1M monthly unique web visitors worldwide in May 2026, the most of any assistant, counting the web product only and excluding its iOS and Android apps.
Which AI model is the most capable in 2026? On GPQA Diamond, Gemini 3.1 Pro leads the consumer flagships at 94.3%, with GPT-5.5 and Claude Opus 4.8 at 93.6%; on SWE-bench Verified, GPT-5.5 (88.7%) and Claude Opus 4.8 (88.6%) lead.
Which AI chatbot is most used in the United States? ChatGPT is the most-used AI chatbot in the United States, at 58.3% web-visit share in May 2026, ahead of Gemini (19.3%) and Claude (13.4%).
How often is this report updated? Monthly, using the latest complete month of Similarweb data.
Methodology and data sources
- Metric: web visits (a session on the app's primary web domain), and visit-derived web-visit share within this defined group of seven assistants.
- Data source: Similarweb, total = desktop and mobile web. Frontier-model capability and pricing data from the DemandSphere AI Frontier Model Tracker, used with permission (June 2026 report).
- Data window: May 2026 (latest complete month). Trends run February 2025 through May 2026. Month-over-month compares April 2026; quarterly compares February 2026; six-month compares November 2025; year-over-year compares May 2025.
- Scope of data: worldwide headline, a United States table, a 15-month trend, per-country breakdowns for the top three apps, a US platform breakdown of web versus app users, and a frontier-model capability comparison.
- Inclusion rule: standalone, general-purpose generative AI assistants on a separable primary web domain above roughly 50M monthly web visits.
- Excluded apps: assistants without a separable domain (Meta AI), companion and character apps (Character.AI), aggregators (Poe), major non-English assistants (Doubao, Ernie, Qwen), and small standalone apps (Brave Leo, Komo, Andi).
- Known limits: web visits exclude native apps, embedded surfaces, and API or enterprise usage; some apps' web visits may be inflated by ecosystem entry points; the platform and model tables use different metrics and time bases from the web-visit ranking; estimates are panel-modeled, carry error bands, and are revised monthly.
- Update cadence: monthly.
