OpenAI Publishes ChatGPT Signals Data On Work Activities

OpenAI used its newsroom on October 26, 2026, to publish a page built around a dataset it calls Signals, which the company says tracks 165 distinct work activities as they appear in ChatGPT conversations across nearly 100 countries. Rather than announcing a new model or a pricing change, the page is framed as a way to let outsiders look at how people actually put the assistant to work, and how that pattern moves as adoption spreads and matures in different economies.

The canonical page at openai.com/index returned a block to the automated reading used for this site, which is a known limitation on pages under that path. The text summarized here comes instead from OpenAI’s own official news feed entry describing the same release, which still names the page, the event, and the scope of the data behind it.

The company frames the release as evidence that ChatGPT’s role at work is still shifting rather than settled. By tracking the same 165 activities over time and across a wide country set, OpenAI is positioning Signals as a way to show trends rather than a single snapshot, the kind of claim that is hard to make credible without a large, repeated sample. Work activities in this context mean the kinds of tasks a person brings to the assistant during a job, not categories OpenAI invented for marketing purposes, and the breadth of the country coverage suggests the company is trying to speak to a global customer base rather than just the US market where much of its subscriber growth has concentrated.

The numbers

OpenAI’s own description names two scope figures for this release: 165 work activities tracked, and coverage across nearly 100 countries. Those two figures describe the breadth of the Signals dataset rather than a usage total, a revenue figure or a subscriber count, so this release should be read as a measurement exercise rather than a business update.

For a reader deciding how much weight to put on vendor-published usage research, the relevant question is less about the headline figures and more about what counts as a work activity in OpenAI’s own taxonomy, since that categorization choice shapes every later comparison OpenAI or anyone else draws from the data. No independent outlet has reported on this release at the time of writing, so this account rests on OpenAI’s own description of what it published and why.