A program of EffectiveSC
How We Use AISources checked October 24, 2026 to November 2, 2026
Where things stand / Which AI tools people use, and how
Which AI tools lead, who uses each one, what it is used for, and how people reach it: inside other apps, on a phone or a computer, or by voice.
Robb Wilson, who leads this research, works at OneReach.ai, which sells agent orchestration software, and EffectiveSC, which publishes How We Use AI, plans to offer AI services of its own. Claude, Anthropic's AI, helps research this site, and Claude is one of the tools compared here. Menlo Ventures, whose surveys are cited, invests in Anthropic. Ramp, whose customer data is cited, sells corporate cards and software for managing spending, including spending on AI. OpenAI, Anthropic, Microsoft and Google each studied logs of their own products, and those cards say so.
See also: Who publishes How We Use AI, What else we plan to sell or offer
Among American adults, ChatGPT leads by a wide margin, then Gemini, Copilot and Meta AI, with Claude far behind, and Korea's official survey puts ChatGPT even further ahead. Among people who already use AI, one investor's survey finds Gemini has nearly caught up. Among businesses that pay for AI through one spending platform, Anthropic and OpenAI are close, and Anthropic takes most of the spending on tokens. Who leads depends on who is counted: everyone, people who use AI, or companies that pay for it.
44%
In February 2026, 44% of US adults said they ever use ChatGPT, 24% Gemini, 17% Copilot, 14% Meta AI, 8% Grok, 6% Claude and 3% Character.ai.
Who was asked: 5,119 US adults, Pew Research Center's American Trends Panel (probability-based, weighted)
ChatGPT was asked of everyone, and each other chatbot of one random half of the sample. Ever using a chatbot is not using it often.
Disclosure: Claude is Anthropic's product, and Claude, Anthropic's AI, helped compile this research.
31%
In a March 2026 probability-panel poll, 31% of US adults had used ChatGPT in the past week, 21% Google Gemini, 11% Microsoft Copilot, 8% Meta AI, 5% Grok, 3% Claude and 2% Perplexity; 49% had used no AI service that week.
Who was asked: 2,021 US adults aged 18 or older, Ipsos KnowledgePanel (probability-based, weighted)
Use in the past week, a stricter test than ever having used a tool, and more than one service could be chosen.
Disclosure: Claude is Anthropic's product, and Claude, Anthropic's AI, helped compile this research.
60%
In Menlo Ventures' July 2026 survey, 60% of US AI users used ChatGPT, 58% Gemini and 20% Claude, up from 45%, 36% and 7% in 2025; the average AI user used 3.0 general AI assistants, up from 2.2.
Who was asked: AI users among 5,067 US adults, online survey with Morning Consult, July 2026
Shares of people who use AI, not of all adults, from an online panel. Most use more than one assistant, so the shares add up to more than the whole.
Disclosure: Menlo Ventures invests in Anthropic and marks Claude as one of its portfolio companies in this report; Claude, Anthropic's AI, helped compile this research.
41.8%
In Korea's 2025 internet-use survey, 44.5% of people aged 3 and older had used a generative-AI service, up from 33.3% a year earlier; the services named were ChatGPT (41.8%), Gemini (9.8%), Copilot (2.2%) and CLOVA X (2.0%), with more than one answer allowed. Office workers had the highest rate of use, 71.9%.
Who was asked: 50,750 people aged 3 and older in 22,671 households across Korea
Korea's official survey of the whole population, not of workers. More than one service could be named, and the release doesn't say whether the service shares are of everyone or only of people who had used generative AI.
43.8%
In September 2026, 55.9% of US businesses on Ramp paid for AI; 43.8% paid Anthropic and 40.6% paid OpenAI.
Businesses that pay for AI through one spending platform, not all businesses. Free tools and personal accounts are missed.
Disclosure: Ramp sells corporate cards and spend-management software, including a product for managing spending on AI, and these are its own customers.
60.5%
Among businesses using Ramp's AI token spend product, Anthropic took 60.5% of token spending and OpenAI 35.1% in the week ending October 23, 2026, against 51.3% and 44.2% in the week ending September 25; Anthropic had 59.8% of token volume.
Spending on tokens by the Ramp customers who use its token product, which Ramp says doesn't represent all its customers' spending on AI. The leader's share moved by several points in four weeks.
Disclosure: Ramp sells the token spend-management product these figures come from, and these are its own customers.
40%
Foundation-model API spend was $12.5 billion in 2025; by estimated spend, Anthropic had 40%, OpenAI 27% and Google 21%. Open-source models' enterprise share fell to 11%, a separate measure.
Menlo's estimate, from a survey of enterprise AI decision-makers and its own sizing of the market. The open-source share is a separate measure.
Disclosure: Menlo Ventures invests in Anthropic.
See also: Old tools or new
Adults sixty-five and older use every chatbot least. Men and women use ChatGPT equally, but more men use Gemini, Copilot, Grok and Claude, and ChatGPT's early lean toward men is gone by its maker's count. In a large Danish study that compares men and women in the same jobs, women were less likely to have used ChatGPT for work. Claude's users lean toward professionals who pay, by one investor's account, and Claude is used more in richer countries. Among businesses on one spending platform, information and finance firms pay for AI most often, and small firms least.
61%
In February 2026, 61% of US adults ages 18 to 29 said they ever use ChatGPT, against 19% of those 65 and older; for Gemini the shares were 30% and 11%, for Copilot 17% and 10%, and for Claude 10% and 1%.
Who was asked: 5,119 US adults, Pew Research Center's American Trends Panel (probability-based, weighted)
Each chatbot but ChatGPT was asked of half the sample, so the age groups are small and some gaps may not be significant.
Disclosure: Claude is Anthropic's product, and Claude, Anthropic's AI, helped compile this research.
22%
In February 2026, the same share of US men and women said they ever use ChatGPT (44% each), but men were more likely to use Gemini (29% vs 20%), Copilot (22% vs 13%), Grok (11% vs 4%) and Claude (9% vs 4%); 13% of men and 15% of women used Meta AI.
Who was asked: 5,119 US adults, Pew Research Center's American Trends Panel (probability-based, weighted)
Ever used, not how often: Pew also finds men more likely than women to use chatbots every day. Each chatbot but ChatGPT was asked of half the sample.
Disclosure: Claude is Anthropic's product, and Claude, Anthropic's AI, helped compile this research.
48%
In OpenAI's study of consumer ChatGPT, about 80% of weekly active users with gendered first names had typically masculine names in the first months after launch, falling to 48% by June 2025; among users who gave their age, about 46% of messages came from those 18 to 25.
Gender is inferred from first names, leaving out names not clearly tied to one gender, and age comes from users who gave it. Consumer plans only.
Disclosure: OpenAI's researchers studied OpenAI's own product.
16 percentage points
In a survey of about 18,000 workers in 11 AI-exposed Danish occupations, run from November 2023 to January 2024, women were about 16 percentage points less likely than men in the same occupation to have used ChatGPT for work, and 12 points less likely when compared with men in the same workplace with the same mix of tasks; each year of age went with a 0.7 percentage point lower likelihood. 41% of the workers had used ChatGPT for work.
Who was asked: About 18,000 valid responses from 100,000 workers invited in 11 AI-exposed occupations in Denmark
Eleven occupations heavily exposed to AI, in one country, at the end of ChatGPT's first year. The measure is having used it for work; the working-paper version reported a larger gap on a broader measure.
82%
In Menlo Ventures' July 2026 survey, Claude's users were more likely than ChatGPT's or Gemini's to be tech or finance professionals with an advanced degree; 82% of them paid for AI, 48% ran AI agents, 1.5x the rate among ChatGPT or Gemini users, and they were twice as likely to spend $100 or more a month on AI. Gemini's users were a few years older than ChatGPT's and otherwise nearly the same.
Who was asked: AI users among 5,067 US adults, online survey with Morning Consult, July 2026
Menlo describes Claude's users against ChatGPT's and Gemini's without printing every comparison figure, from an online panel of people who use AI.
Disclosure: Menlo Ventures invests in Anthropic and marks Claude as one of its portfolio companies; Claude, Anthropic's AI, helped compile this research.
0.7%
Across countries, a 1% higher GDP per working-age person went with 0.7% more Claude.ai use per working-age person in Anthropic's August 2025 sample; across US states the rise was 1.8% per 1% of GDP per person, though income explained less of the variation there.
A correlation across countries in Claude's consumer chat app, not a cause. Income explains less of the gap between US states.
Disclosure: Anthropic studied its own product, and Claude, Anthropic's AI, helped compile this research.
79.9%
In September 2026, 79.9% of information-sector businesses on Ramp paid for AI and 73.7% of finance and insurance businesses, against 42.5% in healthcare and 33.0% in accommodation and food; 65.7% of large, 62.5% of medium-sized and 50.0% of small businesses paid for AI.
Ramp's customers that pay for AI, sorted by Ramp's own models of sector and size. Ramp doesn't say which vendors each sector or size pays, and free tools and personal accounts are missed.
Disclosure: Ramp sells corporate cards and spend-management software, including a product for managing spending on AI, and these are its own customers.
See also: Who is more likely to use it
The product decides how much of the use is work: a minority on consumer chat apps, most of it in workplace tools and through the API. The same assistant is a work tool on a desktop in office hours and a health adviser on a phone. People reach assistants through other apps as well as their own: Meta AI mostly inside messaging and social apps, Gemini in search results and other Google products, Copilot inside other Microsoft products. Phones and computers are used about equally. Voice is a minority habit, and how big a minority depends on how the question is asked.
82% to 89%
In a May 2026 paper, Microsoft estimated that 82% to 89% of Microsoft 365 Copilot Chat use in enterprises is for work and about 10% personal, against 27% work use on consumer ChatGPT in OpenAI's study, so the share of use that is work depends on the product.
Microsoft's estimate for the chat in its office suite at companies, beside OpenAI's figure for consumer ChatGPT; the two are not one measure. Copilot inside Word, Excel and Outlook isn't counted.
Disclosure: Microsoft's researchers studied Microsoft's own product.
86%
In April 2026 logs, over 86% of interactions with Google's conversational AI, the Gemini app and AI Mode, happened outside work.
The Gemini app and Google's AI Mode for consumers; enterprise and Workspace use are excluded.
Disclosure: Google's researchers studied Google's own products.
74%
In November 2025, 74% of sampled requests to Anthropic's API, mostly from businesses, were work-related, against 46% of conversations on its Claude.ai chat app.
Two samples from the same company: requests to its API, mostly from businesses, and conversations in its consumer chat app.
Disclosure: Anthropic studied its own product, and Claude, Anthropic's AI, helped compile this research.
In 37.5 million consumer Copilot conversations from January to September 2025, health and fitness was the top topic on mobile at every hour of the day, while on desktop work and career overtook technology as the top topic between 8 a.m. and 5 p.m.
Ranks of topics, not shares: work leads on a desktop in office hours without being most of the use. Consumer Copilot only, with work and school accounts excluded.
Disclosure: Microsoft's researchers studied Microsoft's own product.
56%
In Menlo Ventures' July 2026 survey, 56% of US AI users used AI weekly in phone apps and 54% on computers, and 15% used it weekly on wearables.
Who was asked: AI users among 5,067 US adults, online survey with Morning Consult, July 2026
Shares of people who use AI, from an online panel. Many use AI on more than one device.
Disclosure: Menlo Ventures invests in consumer AI companies, including Anthropic.
55%
In March 2026, among US adults who had used each in the past week, 55% of Meta AI users reached it inside WhatsApp, Instagram, Facebook or Messenger and 21% through its own app or site; 39% of Gemini users typed into its app or site and 32% used it inside another Google product such as Gmail or Docs; 42% of Copilot users used its app or site.
Who was asked: US adults who used each service in the past week, from 2,021 US adults aged 18 or older
Each share is of that service's own users in the past week. The groups are small, so their margins are wider, and more than one way could be chosen.
15%
In March 2026, among US adults who had used an AI service in the past week, 27% had ever used AI inside another app, such as Copilot in Word or Gemini in Gmail, 15% had had a voice conversation with an AI service, and 9% had used one that completed tasks on its own.
Who was asked: US adults who used an AI service in the past week (n = 1,032, unweighted)
Adults who used AI in the past week and had ever had a voice conversation with it, a narrower test than speaking a prompt; a broader question, on the next card, finds far more voice use.
46%
In Menlo Ventures' July 2026 survey, 19% of US AI users said voice was their main way of interacting with AI and another 27% moved between voice and typing, so 46% spoke to AI at least some of the time, mostly inside AI apps through voice modes, dictation and spoken prompts; 25% of AI users 65 and older counted voice among the ways they use AI most, against 23% of Gen Z.
Who was asked: AI users among 5,067 US adults, online survey with Morning Consult, July 2026
Voice here includes dictation and spoken prompts inside chat apps, a broader test than a voice conversation. The question is new, with no earlier year to compare.
Disclosure: Menlo Ventures invests in Anthropic and in Wispr Flow, a dictation app its report praises in the same passage.
29.8%
Among Japanese small and medium firms that had introduced AI, 82.6% used generative AI and 29.8% speech recognition or spoken-dialogue AI, such as minute-taking or automated answering; 11.2% used image recognition. 20.4% of the firms surveyed had introduced AI.
Who was asked: 322 firms that had introduced AI, from 1,647 responses by 10,000 small and medium firms invited nationwide
Firms, not people: shares of the small and medium firms that had introduced AI, in a government agency's web survey with a low response rate.
See also: What people do with it, What the logs show, and what people say
No survey found asks how many people talk to ChatGPT, Gemini or Claude by voice, or what they use voice for, and the two surveys of people that ask about voice at all ask different questions. Nothing published compares AI agents that answer customers by phone with those in web chat. Ramp splits paying for AI by sector and company size, but not which vendor each sector or size pays. Pew doesn't publish chatbot brands by income, education or job. No source measures use inside other apps across every product, and each company's logs cover its own products, so their shares don't add up to a market.
The AI at Work Benchmark
Score your organization's practices, measure your people with an anonymous pulse, and see both against the published evidence.