OpenAI’s September 2026 data classified 20.2% of sampled messages from individual ChatGPT accounts as Seeking information, tracking message topics rather than how many people searched online.
OpenAI’s Signals data lists a 16.7% share in July 2024, making the September figure 3.5 percentage points higher. The series covers July 2024 through September 2026, and OpenAI updated the data Oct. 26. Organizational usage appears in a separate series.
The unit classified is an individual message. The data dictionary groups requests for specific information, requests about purchasable products, and cooking or recipe prompts under this label. Because the category includes those different subjects, the 20.2% figure reflects the sample’s topic mix; it is not a count of how many people asked questions. It sorts message subjects rather than identifying which product feature someone used.
OpenAI’s methodology sets the monthly consumer sample at one million messages. It draws from ChatGPT Free, Go, Plus and Pro accounts whose users declared an age of at least 18. Deleted accounts, users who opted out of training and Codex messages are excluded, according to the methodology.
These rules define what the percentage covers. It is not an all-accounts total: the published individual sample is drawn from the named plans and age-qualified accounts, while organizational use is reported separately. The exclusions further narrow the population covered by the statistic.
The Signals topic share also does not tell whether a person used the platform’s Search function, whether an answer drew on the internet, or whether the exchange led to an outside website visit. Nor can this classification establish ChatGPT’s portion of Google search activity or show that anyone shifted a query from Google to ChatGPT.
Changes in the percentage can be compared across the series as topic shares within OpenAI’s defined sample. The latest month available in the update is September 2026.
