This article was drafted together with AI (Claude) and reviewed by an editor. Claude is among the products being compared, so it has a stake. For that reason we kept the numbers in the rankings to a minimum and noted where each figure comes from. Some of the figures below were confirmed not from the research firm's original but through the outlets that cited it.
"Which AI do people use the most these days?" is the most common question, and the answer differs from survey to survey. Looking at the four we found, the reason becomes clear.
The Same Question, Four Answers
Below are figures for ChatGPT, Gemini and Claude, in that order. Each number differs in what was counted.
- US web and mobile monthly users (First Page Sage, August 2026): 51.5%, 27.6%, 10.2%
- App usage (Sensor Tower, end of May 2026): 46.4%, 27.7%, 10.3%
- Web visits (based on Similarweb, cited by Momentic, May 2026): 53.9%, 27.4%, 8.2%
- Web traffic (StatCounter, June 2026): 76.87%, 7.94%, 3.74% — here Perplexity is at 7.91%, almost the same as Gemini.
Why the Numbers Diverge
The first three surveys paint a similar picture. ChatGPT in first place is around half, Gemini in second is a little over a quarter, and Claude ranges from a single digit to a little over 10%. The last survey, by contrast, is completely different. The reason is simple. If you count only visits to websites, phone apps and AI built into other services (such as Gemini inside Google Search) drop out. How to count one person using one AI ten times a day versus using three AIs once each also differs from survey to survey. And some of the four surveys look only at the US.
"Used a Lot" and "Liked" Are Different
Heavy use mostly means it became known first and is easy to access. Preference has to be read from other data. On the Arena ranking (the page as of October 2), where people vote for the better of two anonymized answers, Google's Gemini 4 Argon is shown in first place in the text category, followed by three Anthropic Claude models and OpenAI's GPT-6 models. In the development work (WebDev) category, Claude Sonnet 5.5 and GPT-6.1 Sol appeared in third and fourth. However, we did not view the whole page ourselves but read a summary, and the score notation was inconsistent, so we carry over the rankings only. These rankings also change every time a new model is released.
Our Interpretation
Fact: All three surveys agree that ChatGPT leads in share of use, followed by Gemini and Claude. In preference votes, the order came out different from the share of use.
Interpretation: There is little reason to pick a tool by following "the mainstream." A large number of users may be the result of marketing and accessibility. Conversely, being ranked first is no guarantee that it suits your work.
What we would do: We treat share of use only as a reference meaning "other people use this tool too, so it is easy to find materials and examples." We decide the criteria for choosing based on the results of trying it ourselves on three tasks we do every week. The method is in the next article.
Further reading
All figures are from external surveys; where we could not open the source data ourselves, they are as reported by the citing outlets.