OpenAI has published a study of how scientists use ChatGPT, alongside expanding its free researcher-access programme. Participants estimated that equivalent work without AI would take about 1.9 times as long, but the study does not establish that AI caused faster or better research.
OpenAI Watch analysis
What happened
The study analysed 580,186 scientific messages from 1,819 scientists between 20 August and 22 September 2026, and surveyed 251 scientists from 30 September to 7 October. OpenAI calculated the 1.9-times estimate across eight research activities, giving each activity equal weight. When respondents said they did not use AI for an activity, the calculation treated the time as unchanged.
Separately, 98% of the 250 scientists who answered a question about discovery timelines expected AI to accelerate discovery in their field. CASRAI’s account of OpenAI’s announcement says the programme aims to reach 200,000 researchers worldwide. Applications for 50,000 new grants open on 23 November, with the first eligible applicants selected first come, first served. New grants provide Plus access for 12 months to the recipient and up to two collaborators. API credits and specialised models are not included.
Why it matters
The study offers a snapshot of scientists’ reported use and expectations, not a controlled measure of productivity or scientific breakthroughs. OpenAI says the findings do not establish causal effects, and the discovery figure records what participants expect rather than what has been demonstrated. Useful evidence, then, but not quite a lab coat for a sweeping claim.
The programme details are more immediately practical for research teams: the next intake has a dated opening, a 50,000-grant limit and a specific Plus plan. Teams should check their eligibility and whether the included access fits their work before relying on it; OpenAI’s guide says final eligibility and collaborator rules for the new intake remain to be confirmed.
Our read
The 1.9-times estimate is worth discussing precisely because it is self-reported, not because it settles the productivity debate. OpenAI’s findings show what participating scientists say they experience; measuring whether AI improves research outcomes is a different and harder exercise. If you are considering the grant, note the 23 November opening and the limits: Plus access, up to three people, 12 months, no API credits.
What to watch
- Whether later evaluations measure research output or quality, rather than relying only on participants’ estimates.
- Which eligibility and collaborator rules OpenAI confirms for the new intake.
- Whether the 50,000 grants fill quickly after applications open.
Discussion spark: Should self-reported time savings count as persuasive evidence for AI’s value in research, or should funders wait for measured improvements in research output?
Sources and evidence
- OpenAI expands free AI access to 200,000 researchers – CASRAI (16 November 2026, 18:09 UTC)
OpenAI Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by OpenAI.