OpenAI's Economic Research team looked at how scientists actually use ChatGPT, studying more than 580,000 scientific messages from 1,819 scientists in its ChatGPT for Academic Researchers summer pilot and surveying 251 participants.

The researchers used the tool across the whole process, from reviewing literature and analysing evidence to developing ideas and writing code. Those surveyed estimated that producing the same work at the same quality would have taken 1.9 times as long without AI, and 98% of those who answered said they expect AI to accelerate discovery in their field.

Idea work was a big part of it: 68% of the scientists used ChatGPT to develop research ideas. About a third of the time they arrived with a specific question, a similar share brought an idea to refine together, and in about a quarter of cases they asked ChatGPT to suggest a question.

When asked why discovery could move faster, the most common answer was simply getting research done sooner, followed by finding more promising questions and pursuing a wider range of them. OpenAI says it is expanding the academic researchers programme, with free access for more than 200,000 researchers worldwide.

The study was designed around privacy: by default OpenAI does not use participants' programme data to train its models, and the team sorted messages with automated classifiers rather than reading the messages or responses themselves.