Information seeking accounted for 20.7% of messages on consumer ChatGPT plans in August, nearly matching writing at 20.6%, according to OpenAI Signals data released Oct. 26. It was the first month in the series, which starts in July 2024, when the reported share for seeking information was higher. Writing moved back in front in September, at 21.4% versus 20.2%.
In addition to publishing the new version of their signals (Signals V3.0), OpenAI posted a blog post about how users are bringing work to ChatGPT. Their data is based on a monthly sample of 1M messages from account holders with a declared age of 18+ on Free, Go, Plus, or Pro plans. They exclude Enterprise and Codex. In addition, they have created an automated system that classifies each sampled message by subject area.
Over the months both cover, Signals also shows a smaller rise in seeking information than a 2025 paper by OpenAI and Harvard researchers, which put the category at 24% of ChatGPT usage by mid-2025.
Seeking Information Rose While Writing Fell
Writing fell from 34.1% of messages in July 2024 to 21.4% in September 2026. Seeking information rose from 16.7% to 20.2% over the same months.
Practical guidance, which includes how-to advice and tutoring, has been the largest topic every month since March 2025 and was 28.1% in September.
According to the release’s data dictionary, seeking information covers requests for facts, questions about products and services for sale, and recipe or cooking help.
Work-related messages made up 32.2% of all messages in September. Among these, the share of messages seeking information remained between 10.7% and 12.1% each month from July 2024 through July 2026, before rising to 13.9% in August and 15.2% in September.
What The Work Activity Post Adds
The work activity post organizes work-related messages into 165 activities from O*NET, a standard framework that groups tasks shared across jobs. It presents the data by quarter, covering Q1 2025 through Q3 2026, across nearly 100 countries.
Across the 74 countries with data in every quarter, the 10 most common activities averaged 51.1% of each country’s classified work messages in Q3 2026, down from 56.0% in Q1 2025. The post describes that as work-related use becoming less concentrated in the most common activities.
The post’s graph shows “Getting information,” which is one of these 165 activities. In the United States, it represented 11.6% of work messages during the first quarter of 2025. It also showed that it made up 9.9% of work messages in the third quarter of 2026. These are based on a separate classification system and represent only one country (the U.S.), so they do not measure the same thing as “seeking information” globally.
How It Compares With The 2025 Paper
Last September, I covered the OpenAI and Harvard paper on how people use ChatGPT. It reported that seeking information grew from 14% to 24% of ChatGPT usage in the year after July 2024.
Signals utilizes the exact same classifier prompts used by the research study, according to the provided data dictionary, and both categorize the same three categories of requests under “seeking information.” Signals’ graphs also show the percentages of the same time frame for seeking information; 16.7% in July 2024 and 18.3% in June 2025, which is the final month covered by the research paper.
The two writing series have fewer differences between them than the two seeking-information series. The percentage decreased by 12 points (from 36% to 24%) in the research paper while it decreased by 8.4 points (from 34.1% to 25.7%) in Signals during the same time frame.
Approximately 1.1 million conversations were selected by the researchers using the sampling method where they took only one message from each conversation. Each share was then averaged over 28-day periods. Signals selects 1 million messages per month. This release altered previous months to be consistent with that higher number. No reason was stated or explained within the release blog or the provided data dictionary regarding the difference in the seeking information percentages.
Why This Matters
In OpenAI’s own monthly data on consumer ChatGPT messages, approximately twenty percent of all messages are to look something up, which is roughly equivalent to the percentage for “writing.” Product questions are included in this percentage, however because OpenAI combined all three categories (fact/recipe/product) into one category, it cannot be determined how many were each.
Signals does track the various topics of conversation that occur via ChatGPT; however it did not provide any information regarding how often ChatGPT accessed external sources such as websites or cited a site in its replies.
When citing percentages representing the use of ChatGPT, please reference the source and time frame. There was an almost six point difference between the 2025 paper and Signals regarding information seeking in mid-2025.
Looking Ahead
So far, each update in the Signals data dictionary’s changelog adds a quarter of data and revises earlier months. Because a figure from v3.0 may differ from the same month’s figure in a later version, be sure to note the version number whenever you include a Signals figure in a report.
Featured Image: Tetiana Yurchenko/Shutterstock
