GEO WIKI · USERS & USAGE

Claude (Anthropic): Users and Usage 2026

Reach and observed usage (Anthropic Economic Index), for the U.S. and worldwide.

· Last reviewed: August 8, 2026 · First published: August 8, 2026

Claude 2026: Where things stand

Reach (as far as reliable) and the U.S. market from the Anthropic Economic Index.

Worldwide (reach)
~ $30B
Revenue run-rate
annualized, global, 2026·global·2026
Context
> 2M
Claude Code (WAU)
weekly active users, global, May 2026·global·May 2026
Official
Source: Anthropic
≈ 245M
Monthly users
estimate, global, mid-2026·global·mid-2026
Estimate
Source: Sensor Tower

How to read these three figures: the run-rate is a revenue number Anthropic stated publicly and a secondary source reported, not a user count. Claude Code WAU counts only the weekly users of the Claude Code developer tool, not Claude overall. The 245 million monthly users is a third-party estimate; Anthropic itself publishes no such figure. So there is no official total Claude user number.

United States (measured from Anthropic data)
3.87
Anthropic Usage Index
rank 12 of 121 countries, May 2026·United States·May 2026 (snapshot)
Official

The Anthropic Usage Index compares a country's share of Claude usage with its share of the working-age population. 1.0 is the global average. The value 3.87 for the U.S. means about 3.87× the Claude usage that population size alone would predict. „Measured from Anthropic data" means this comes from actual, anonymized usage, not from a survey.

Reach is not behavioral usage

Two different questions. Reach: how many people use Claude? Anthropic discloses little here; the reliable figures are revenue run-rate and Claude Code WAU, with consumer MAU only as a third-party estimate. Behavioral usage: what is Claude actually used for? That is what the Anthropic Economic Index measures from anonymized, aggregated conversations. The two levels are kept apart.

What the Economic Index is NOT

The index describes observed Claude usage by task, not the labor market, not jobs, and not who the users are. Shares always refer to classified Claude conversations, never to a „share of people“. It is a snapshot (data period May 2026), not a time series, and not evidence of job automation. Source and methodology: anthropic.com/economic-index.

Quick definition · continuously maintained reference (Users & Usage)

This continuously updated reference covers Anthropic's Claude on two separate levels. First, reach (how many people use Claude), which Anthropic discloses only sparingly. Second, observed usage: the Anthropic Economic Index measures, from anonymized and aggregated Claude conversations, what tasks Claude is used for, worldwide and by country. Reach and behavioral usage are never merged into a single figure.

Reach of Claude: what Anthropic (does not) publish

Anthropic discloses user numbers sparingly. What is reliable and what is not.

Unlike Google (Gemini app MAU) or OpenAI (ChatGPT „weekly active users“), Anthropic reports no regular consumer reach for Claude. That is the data gap: the three figures above measure three different things, and none of them is an official Claude MAU.

Reading the reach

The run-rate is a revenue figure, not a user metric, and comes from a secondary source (a reported Anthropic figure). Claude Code WAU covers only the developer tool, not Claude overall. The circulating MAU (~245M) is an external panel/market estimate, not equivalent to the official, precisely worded reach figures from Gemini or ChatGPT. So the reliable signal for Claude sits on the usage level (Anthropic Economic Index), not on reach. The rest of this page therefore rests on the usage analysis.

What Claude is actually used for (Anthropic Economic Index)

Worldwide usage by task, topic and collaboration style. Data period May 2026 (snapshot).

The Anthropic Economic Index classifies anonymized, aggregated Claude conversations by which occupational tasks and topics they match. The shares below refer to classified Claude conversations worldwide, not to people.

Collaboration & context (worldwide)
51% / 49%
Augmentation vs. automation
user stays involved vs. Claude completes the task·global·May 2026
Official
43% / 40% / 16%
Work / personal / coursework
share of classified conversations·global·May 2026
Official
~ 5h → ~ 40min
Estimated time per task
alone vs. with Claude, rough order of magnitude·global·May 2026
Estimate

In plain terms: augmentation means the person stays involved and works step by step with Claude. Automation means Claude completes larger parts on its own. For Claude the two are roughly balanced. The second tile shows that work and personal use are almost even, with study and coursework about one sixth. The time figure is a rough classifier estimate, not a measured duration.

The next two tables show which job categories and topics Claude is used for most worldwide. The percentages are shares of all classified Claude conversations, not shares of people.

Top job categories by share of classified Claude usage worldwide
Job categoryShare (worldwide)
Computer and Mathematical23.8%
Arts, Design, Media13.6%
Educational Instruction12.8%
Sales9.1%
Office and Administrative7.9%
Management5.9%
Business and Financial5.8%
Top request topics by share of classified Claude usage worldwide
Request topicShare (worldwide)
Content Creation & Copywriting22.7%
Education & Learning13.2%
Software Development11.5%
Research & Intelligence10.9%
Hobbies & Lifestyle9.5%

Key takeaway: the largest block is computer- and mathematics-related tasks (about a quarter), followed by creative and educational tasks. Among topics, writing text (content and copywriting) leads, ahead of learning and software development.

How to read this

The correct reading is „Claude is used for tasks commonly done in job category X“, not „people in job X use Claude“. The people in these conversations are often not members of that occupation. A snapshot here means a single point in time (data period May 2026), with no trend over time.

How the U.S. uses Claude

The U.S. usage profile against the global average. Anthropic-measured.

The United States has an Anthropic Usage Index of 3.87 (rank 12 of 121 covered countries). The index is a country's share of Claude usage divided by its share of working-age population: 1.0 is the global average, so 3.87 is 3.87× what population share alone would predict. The U.S. profile differs from the global average in a few clear ways.

41% / 50% / 9%
Work / personal / coursework
United States, May 2026·United States·May 2026
Official
50% / 50%
Augmentation vs. automation
United States, May 2026·United States·May 2026
Official
rank 12 / 121
Country ranking
by Anthropic Usage Index·United States·May 2026
Official

Two differences from the global average stand out. First, personal use leads in the U.S. at 50 percent, while work and personal are almost even worldwide (43 to 40 percent); the work share is therefore lower in the U.S. Second, coursework is well below the global level (9 vs. 16 percent). In the country ranking the U.S. sits at rank 12 of 121.

The following table shows which job categories Claude is used for relatively more or less in the U.S. than worldwide. A value of 1.0 equals the global average: above 1.0 is over-represented, below 1.0 is under-represented.

United States against the global average by job category
Job categoryU.S. index (1.0 = global)Read
Healthcare Practitioners1.31×over
Sales1.27×over
Community and Social Service1.25×over
Business and Financial1.09×slightly over
Computer and Mathematical0.89×under

What stands out: Claude is used above average for tasks in healthcare, sales and community/social service, while computer- and mathematics-related work sits slightly below the global average. This does not mean those occupations use Claude; it means the conversations match tasks typically found in those fields.

U.S. topics

On request topics the U.S. runs above the global average for Hobbies & Lifestyle (13.6% vs. 9.5%) and Research & Intelligence (13.5% vs. 10.9%), and below for Education (9.3% vs. 13.2%) and Software Development (8.1% vs. 11.5%). All shares refer to classified Claude usage in the U.S., not to the population.

How granular is the Economic Index?

The breakdown reaches individual O*NET occupations, more than 700 of them. Two examples with real task data.

The Economic Index is fine-grained enough that even niche occupations show up. It is built on the *ONET occupations**, a public catalog from the U.S. Department of Labor that describes each job's typical tasks in plain language. Conversations are matched to those tasks across more than 700 occupations. One obvious and one surprising example show the range.

Search Marketing Strategists
Rank 67 of 718 occupations, 0.36% of classified Claude usage worldwide. Collaboration: 45% augmentation / 55% automation. Frequently observed tasks: optimize web visibility via search patterns (SEO), collect and analyze web metrics such as conversion and click-through rates, market research on search trends and competitors, report KPIs from digital campaigns.
Clergy
Rank 110 of 718, 0.18% of classified Claude usage worldwide. Collaboration: 60% augmentation / 40% automation. Frequently observed tasks: study and interpret religious laws, doctrines and traditions, plan religious education programs, pastoral counseling, write articles, sermons and talks.

The clergy example shows the breadth of the dataset: a highly specialized, strongly language-based field ranks 110 of 718 worldwide. Claude is not used only for tech and knowledge work. The high augmentation share (60 percent) fits a field where people think and phrase along closely rather than handing tasks off entirely.

Clean denominator

Correct: „0.36% of classified Claude usage worldwide matches tasks attributed to the occupation Search Marketing Strategists." Wrong: „0.36% of people use Claude for search marketing." The index measures conversations by task, not people by job.

Also available by country

Individual occupations are also broken out by country. For Search Marketing Strategists: 0.28% of classified Claude usage in the U.S., 0.31% in Germany, 0.36% worldwide. The country breakdown does exist, each with the same clean denominator.

What the mapping means

The mapping describes *tasks that match ONET occupations**, measured from conversation content. It does not prove that the person in the conversation actually holds that job. The augmentation/automation figures are conversation styles, not statements about jobs or their replaceability.

Three levels of claim, kept separate

So observed Claude usage does not become an overstretched market figure.

1. Anthropic-measured
Claude-specific, actually observed usage from the Anthropic Economic Index. Reliable, but only for Claude and only as a snapshot.
2. Cross-provider comparison
Alignment with OpenAI/NBER (ChatGPT usage studies) and, once available, Google ATLAS. Enables patterns across providers where definitions are compatible.
3. Proxy hypothesis
Only if occupational task distributions behave similarly across general-purpose AI assistants could Claude give an order of magnitude for broader AI usage. This level is a hypothesis and never quietly becomes a metric.

This page reports mainly level 1. Level 2 is added once comparable provider data is available; level 3 stays explicitly labeled as a hypothesis. The planned comparison article „How people really use AI: ChatGPT vs. Claude vs. Gemini“ builds on exactly this separation.

Methodology and evidence tiers

Denominator doctrine

Every behavioral figure names its denominator: „share of classified Claude usage (worldwide / in the U.S.)“. Never „share of people“. Reach (users) and behavioral usage (tasks) are kept separate.

The Anthropic Economic Index is licensed CC BY 4.0 and describes observed usage by task, not employment and not automation. It is a snapshot (data period May 2026) with no time series. Source and methodology: anthropic.com/economic-index.

Official
Published directly by Anthropic (Economic Index, company figures, product figures).
Estimated
Third-party estimate or classifier approximation (e.g. MAU panel, time-per-task). Not an exact measurement.
Anthropic Usage Index
A country's usage share ÷ its share of working-age population. 1.0 = global average, 2.0 = twice what population share would predict.
Augmentation / automation
Conversation styles: augmentation = user stays involved; automation = user has Claude complete the task. Not a job outcome.

Change log

  • August 2026 (first publication): reach deliberately marked as thin (run-rate ~$30B company figure; Claude Code >2M WAU; MAU ~245M third-party estimate only). Behavioral module from the Anthropic Economic Index (snapshot May 2026): worldwide augmentation/automation 51/49, work/personal/coursework 43/40/16, top job category Computer & Mathematical 23.8%. United States: Usage Index 3.87 (rank 12/121), personal-led (50%), Healthcare/Sales/Community over-, Computer & Mathematical under-represented. Granularity block with real task data (Search Marketing Strategists: rank 67/718, 0.36% global, 0.28% U.S., 0.31% Germany; Clergy: rank 110/718, 0.18%) and a clean denominator. Three-level separation (Anthropic-measured / cross-provider / proxy hypothesis).

Sources and further documentation

Last reviewed: August 8, 2026