OpenAI has published a new dataset called OpenAI Signals, describing how ChatGPT is used in professional settings. The dataset spans 165 distinct work activities across nearly 100 countries, offering a granular view of actual usage patterns rather than theoretical estimates of AI's potential impact.
This approach differs from OpenAI's earlier publications on the economic impact of generative AI, which relied mostly on projections. Here, the figures reflect observed usage, allowing researchers to track over time how specific tasks — writing, analysis, coding, information retrieval — gain or lose prominence across countries and sectors. By breaking usage down by activity rather than by job title, the dataset helps clarify which specific tasks are being delegated to AI, independent of a user's formal occupation.
This kind of data is of direct interest to economists, labor market researchers, and policymakers trying to gauge the real scale of generative AI adoption, beyond the often speculative discourse around job automation. By releasing these figures, OpenAI is also shaping the public conversation about its own role in reshaping work, a topic on which the company faces regular scrutiny from regulators and labor groups.
It is worth noting that this data comes exclusively from ChatGPT usage and therefore captures only part of the broader generative AI landscape at work, excluding usage through competing tools or internal enterprise deployments. The analytical value of the dataset will also depend on the exact methodology used to classify activities, which OpenAI will need to detail for the research community to use it rigorously.