Claude AI Statistics in 2026: Usage, Adoption, Revenue, and Market Share
A slide that says “Claude has X users” can look tidy and still be wrong. Claude.ai visits, mobile-app monthly active users, API conversations, paid subscriptions, business-spend adoption, and Anthropic revenue all measure different populations.
The short answer: the best Claude AI statistics show substantial adoption across consumer, developer, and business surfaces in 2026, but Anthropic has not published a single consolidated, Claude-wide MAU series in the public research cited here. Use the number that matches your question, and keep estimates, samples, and revenue run rates in their own lanes.
Key Claude AI statistics in 2026
Anthropic sampled 1 million Claude.ai conversations and 1 million first-party API conversations from February 5–12, 2026. This is a two-surface research sample, not a count of all Claude conversations or users. Anthropic’s March 2026 Economic Index is the first-party source.
Computer and Mathematical tasks accounted for 35% of the Claude.ai conversations in Anthropic’s February 2026 sample. The metric describes the task mix within that sample during that week, not all Claude activity.
The 10 most common tasks accounted for 19% of Claude.ai traffic in the February 2026 sample, down from 24% in Anthropic’s November 2025 sample. Anthropic’s earlier research also identified more than 3,000 unique work tasks in Claude.ai. Anthropic’s January 2026 Economic Index provides the November comparison and task-breadth finding.
Anthropic estimated that about 49% of jobs had at least one-quarter of their tasks performed using Claude in its March 2026 methodology. This is a task-exposure estimate, not the percentage of workers with Claude accounts.
Ramp reported Anthropic adoption at 34.4% of businesses in its April 2026 business-spend dataset, versus 32.3% for OpenAI. The measure is based on corporate-card and invoiced payments represented in Ramp’s data; it is not a global Claude customer count. Ramp’s May 2026 AI Index is the first-party source for that dataset.
PPC Land reported Similarweb estimates of 950 million Claude.ai visits and 111 million unique visitors in August 2026, plus 157 million monthly active users for Claude’s mobile apps. These are three separate third-party estimates and must not be added into a single user total. PPC Land’s report of Similarweb estimates is the available source.
Momentic reported a Similarweb estimate of 952.6 million Claude web visits and 9.2% worldwide web-visit share in May 2026. The share covers a defined group of seven AI assistants, not the whole generative-AI market. Momentic’s methodology and estimate are third-party evidence.
First Page Sage estimated Claude at 10.2% of the U.S. generative-AI chatbot market in September 2026. Its model uses estimated monthly active users across U.S. web and mobile platforms, making it a proprietary third-party market estimate rather than an Anthropic disclosure. First Page Sage’s September 2026 estimate explains that lens.
Business Insider reported an Appfigures estimate that Claude’s U.S. downloads rose 240% month over month in February 2026. Appfigures models app-store trends, and downloads are not active users. Business Insider’s Appfigures report is the source for the estimate.
Anthropic said paid subscriptions had more than doubled in 2026, according to TechCrunch. The same report describes Indagari’s analysis of billions of anonymized card transactions from about 28 million U.S. consumers; that panel excludes free users and Anthropic’s enterprise business. TechCrunch’s report therefore measures paid-consumer activity, not Claude’s total audience.
CNBC reported that Anthropic told investors its annualized revenue run rate reached $65 billion at the end of July 2026. CNBC characterized the information as a confidential investor update reported by sources familiar with it, not an audited public filing. CNBC’s August 2026 report is the evidence for the run-rate figure.
Sacra estimated Claude Code at $2.5 billion in annualized revenue in February 2026, with 50% from enterprise. This is an analyst estimate for a product line, not Anthropic-reported company revenue. Sacra’s analysis supplies the estimate.
Get the Mobile Testing Playbook Used by 800+ QA Teams
Discover 50+ battle-tested strategies to catch critical bugs before production and ship 5-star apps faster.
What Anthropic’s official data measures
Anthropic’s Economic Index is the strongest starting point for Claude usage because the company defines what it sampled. Its March 2026 report analyzed 1 million Claude.ai conversations and 1 million first-party API conversations from February 5–12. That design does not produce a global user count. It shows how activity was distributed across two specified product surfaces during a defined period.
The distinction matters because Claude.ai and the first-party API serve different workflows. A Claude.ai conversation is not an API record, and neither is an account, a paying customer, or a monthly active user. Treating those measures as interchangeable would turn a useful usage sample into a false adoption total.
Anthropic found that Computer and Mathematical tasks made up 35% of the February Claude.ai sample. That tells you coding- and math-related work was prominent in the sampled conversations, but it does not establish the share of all Claude tasks across mobile, web, API, Claude Code, and other product experiences.
Anthropic also estimates that about 49% of jobs had at least one-quarter of their tasks performed using Claude. This is a model of task exposure: it describes the tasks associated with occupations in Anthropic’s method. It does not mean 49% of workers use Claude, have accounts, or pay for it.
How Claude usage patterns changed
The clearest first-party change in 2026 is task concentration. Anthropic reported that the top 10 tasks made up 19% of Claude.ai traffic in its February sample, compared with 24% in the November 2025 sample. Its January report’s finding of more than 3,000 unique work tasks adds context: Claude.ai use appeared broad rather than concentrated in a short list of activities.
That decline from 24% to 19% should not be read as a simple story of users independently broadening their use. Anthropic attributes part of the shift to coding work moving from Claude.ai to first-party API traffic as Claude Code grew. The report also recorded coursework falling from 19% to 12% of conversations and personal use rising from 35% to 42%. Product-surface migration and seasonal changes can both affect a task-mix comparison.
Anthropic’s September 2025 Economic Index offers a useful, older view of how the surfaces differed. In its sampled first-party API data, 77% of business uses involved automation patterns, versus about 50% for Claude.ai users. Anthropic’s September 2025 report classifies sampled uses; it does not say that 77% of enterprise customers automate work with Claude.
Trendlines also have a methodology boundary. Anthropic’s June 2026 report says increasingly long-running agentic sessions from Claude Code and Cowork led it to introduce higher-rate sampling, a new output classifier, and more granular monthly reporting that separates Claude/Cowork conversations from first-party API traffic. Anthropic’s June 2026 methodology update means pre-June and post-June results are related evidence, not automatically a perfectly continuous series.
Claude web traffic, mobile activity, and audience estimates
Third-party audience estimates are valuable when their units stay visible. The August 2026 Similarweb figures reported by PPC Land include 950 million Claude.ai visits, 111 million unique visitors, and 157 million mobile-app MAU. A visit is a browsing session. A unique-visitor estimate attempts to identify distinct visitors during a period. Mobile MAU counts active app users under the provider’s model. None is a substitute for the other.
The May 2026 estimate reported by Momentic is a separate web-traffic snapshot: 952.6 million Claude web visits and 9.2% of visits among seven selected AI assistants worldwide. It is not surprising that this estimate differs slightly from the August snapshot. The dates, traffic volumes, and comparison groups differ.
Historical traffic reports also need their own label. SociallyIn reported a Semrush estimate that Claude website visits rose from 16 million in January 2025 to more than 164.8 million in August 2025. SociallyIn’s historical Semrush summary is useful for directional web-traffic context, but those figures are visits rather than users or app MAU.
The public evidence does not support a single official Claude-wide MAU figure. Anthropic’s cited research publishes usage samples and task distributions, while external publishers report web visits, unique visitors, app MAU, or market-share models. A circulating headline number may look more convenient, but combining these measures would double-count people and mix unrelated populations.
Claude mobile downloads and paid-consumer evidence
The 240% month-over-month rise in estimated U.S. downloads during February 2026 is evidence of mobile acquisition momentum. It is not evidence that Claude had 240% more active users. App downloads can include multiple devices for the same person, and a download says nothing by itself about retention or paid conversion.
The paid-consumer evidence is similarly narrower than an all-user estimate. Indagari’s panel covers anonymized credit-card transactions from about 28 million U.S. consumers, not free Claude users, non-card payment behavior, or Anthropic’s enterprise business. Anthropic’s statement that paid subscriptions more than doubled in 2026 is meaningful, but it does not disclose the subscription base or total consumer population.
For a deck or planning model, use these figures to describe consumer momentum: app-download growth, paid-subscription growth, and traffic estimates. Do not use them to infer a verified total of active Claude users.
Claude business adoption indicators
Ramp’s AI Index measures a different form of adoption: businesses with AI-vendor spending visible in its corporate-card and invoiced-payment data. In Ramp’s April 2026 update, Anthropic adoption rose from 24.4% to 30.6% of businesses between February and March. Ramp’s April 2026 update provides the earlier movement under the same spend-based approach.
By the April observation reported in Ramp’s May update, Anthropic reached 34.4% of businesses in Ramp’s dataset, compared with 32.3% for OpenAI. This is strong evidence of business payment adoption in Ramp’s observed population. It is not worldwide enterprise market share, because businesses outside Ramp’s payment data are not part of the measure.
Ramp’s crossover analysis adds context on vendor overlap. It reported that 52% of customers using Anthropic or OpenAI used both, and that 43% of Anthropic customers had switched from another generative-AI vendor. Ramp’s crossover analysis should be read as behavior among customers in its analysis, not behavior among every Claude customer.
The practical takeaway is that business adoption is not a winner-take-all measure. A company can pay for more than one model provider, so vendor adoption percentages cannot be treated as mutually exclusive customer shares.
Claude revenue and market-position evidence
Anthropic’s reported financial traction is substantial, but the metric labels matter. CNBC reported a $65 billion annualized revenue run rate at the end of July 2026, a $47 billion run rate in May 2026, a preliminary $11.5 billion second-quarter revenue figure, and approximately $10 billion in revenue for all of 2025.
An annualized run rate projects a recent revenue pace forward; it is not revenue already recognized across a calendar year. The $11.5 billion figure is preliminary quarterly revenue, while the approximately $10 billion figure refers to the full 2025 year. None should be added to another, and none establishes Claude-only recognized revenue.
Sacra’s $2.5 billion February annualized-revenue estimate for Claude Code is a separate analyst model. It is useful as an indication of how observers size that product line, but it should not be combined with Anthropic’s company-level run rate. The two figures have different publishers, scopes, and evidence types.
Market-share estimates also answer different questions. First Page Sage’s 10.2% September estimate is for U.S. generative-AI chatbot MAU across web and mobile. Momentic’s 9.2% May estimate is worldwide web-visit share within seven assistants. Both suggest Claude was a significant participant in their defined markets; neither is a universal measure of generative-AI usage or revenue.
Why Claude AI statistics disagree
The numbers disagree because they were designed to measure different things, not necessarily because one source is wrong.
Product surface: Claude.ai, mobile apps, Claude Code, and the first-party API generate distinct forms of activity.
Unit of measurement: a conversation, visit, unique visitor, download, active user, paid subscription, and business payment are different units.
Population: Anthropic samples specified activity; Ramp observes businesses in its spend dataset; Indagari observes a U.S. consumer card panel; analytics providers model web or app audiences.
Geography: some estimates are U.S.-only, while others are worldwide.
Time period: a February usage sample, May web estimate, August web estimate, and September market estimate should not be presented as one synchronized snapshot.
Method change: Anthropic’s June 2026 changes to sampling and classification limit direct before-and-after comparisons.
This measurement audit produces a more useful answer than a single headline user number. Claude has documented breadth of use in Anthropic’s samples, large third-party traffic and app-audience estimates, rising paid-consumer signals, and strong business-payment adoption in Ramp’s dataset. The public record does not establish one official, consolidated Claude MAU total.
Methodology and source notes
This report prioritizes Anthropic’s research for Claude usage patterns and methodology, then uses Ramp for its own business-spend index. Web, mobile, and market-position figures are explicitly labelled as estimates from analytics providers or publishers reporting those providers. Financial figures are labelled as journalism based on confidential investor information or as analyst estimates.
The report does not turn sampled conversations into users, visits into MAU, downloads into retained users, or annualized run rate into annual revenue. It also does not reconstruct a Claude revenue total by adding company revenue, Claude Code estimates, mobile activity, and business-spend indicators.
No public source cited here publishes a verified, consolidated Claude-wide MAU series. That gap is important: it tells you to match each statistic to the decision it can actually support rather than forcing every measure into a universal adoption claim.
Conclusion
Claude’s 2026 numbers show expansion across several distinct surfaces, not one neat total. Anthropic’s samples are the right evidence for how work is distributed. Ramp’s index is the right evidence for business payment adoption in its dataset. Web, app, and market estimates are useful for audience context, while revenue run rates belong in financial discussions.
When you use a Claude statistic, lead with the question you need to answer. That choice—not the biggest number available—determines whether your slide, budget case, or competitive brief stays accurate.
Continue reading
More articles from the Quash editorial desk.
