OpenAI Study Finds ChatGPT Users Taking On Tasks Outside Their Jobs
The vendor's analysis suggests AI is changing who performs work before job titles catch up, with task crossover appearing most strongly in several non-engineering roles.
OpenAI says a large share of occupation-specific ChatGPT use involves people doing tasks traditionally associated with another job. Its July 27 Work at the Frontier study analysed more than 800,000 messages from U.S. users and reported that 16.8% of work-related messages, and 43.5% of occupation-specific messages, involved tasks linked to another occupation.
The study calls this pattern task crossover. It offers a useful way to examine how AI may rearrange work before employers formally change job descriptions, but it should not be read as evidence that the tasks were completed accurately, improved productivity or replaced another worker.
How the measure works
OpenAI first separates generic activities such as writing, summarising and scheduling because those appear across too many jobs to demonstrate crossover. It then compares the remaining messages with the user's occupation and the occupation most closely associated with each task.
Among those non-generic messages, the company reports that 43.5% fall outside the user's own occupation. Customer-experience, design, human-resources, legal and marketing workers show especially high outside-occupation shares in the published breakdown. Marketing and engineering tasks also appear widely in messages from people working in other fields.
That does not necessarily mean those occupations are shrinking. It may mean a marketer can troubleshoot a site before asking a developer, or a small-business operator can handle an initial analysis that once required a handoff. Whether the result is efficient expansion, risky overreach or both depends on task difficulty, review and the quality of the model's output.
Small workplaces show a stronger pattern
Among average workspace users, the outside-occupation share falls from 18.9% in workspaces with two to five seats to 16.3% in workspaces with more than 100 seats. OpenAI suggests smaller organisations may have fewer specialists available, while acknowledging that heavy users do not follow the same simple size pattern.
The result is best treated as an early behavioural indicator. The dataset comes from ChatGPT, is analysed by its vendor and covers messages from U.S. users rather than a representative census of all workers. It records what people ask the system to help with, not whether the work succeeds or how employment changes.
Status
Analysis of vendor-published usage research. The sample, method and reported percentages come from OpenAI; the study does not establish productivity gains, job losses or results across the whole economy.
Sources
Update note: Last reviewed 2026-07-28. We will revise this post if OpenAI changes the report, methodology or published figures.
Sources
- OpenAI — official
Drafted with AI assistance from source briefs; reviewed for citation completeness and label accuracy.