Sulat.com / Writing on AITuesday, November 17, 2026
Artificial Intelligence/17 November 2026

OpenAI study: scientists say ChatGPT nearly doubles their research productivity

OpenAI's study of 1,819 scientists has survey respondents estimating the same work would take 1.9 times as long without ChatGPT. The next round of free access for researchers opens on 23 November.

Glossy 3D OpenAI logo in green glass and chrome beside a lab bench with flasks, laptop, microscope and a rising green arrow.

At a glance

  • OpenAI's Economic Research team analysed more than 580,000 messages from 1,819 scientists on the ChatGPT for Academic Researchers plan between 20 August and 22 September 2026, and surveyed 251 of them.
  • Surveyed scientists estimated the same work at the same quality would take 1.9 times as long without AI, with research coding at nearly three times and ideation at nearly twice, OpenAI says.
  • OpenAI says 98% of the 250 scientists who answered expect AI to speed up discovery in their field, and 22% think a discovery that takes 10 years could happen within three.
  • OpenAI says applications for the next round of free ChatGPT Plus grants open on Monday 23 November, with 50,000 grants that each cover a recipient and up to two collaborators.

OpenAI says the scientists in its ChatGPT for Academic Researchers program estimate the same work would have taken 1.9 times as long without AI. The company’s Economic Research team published its report, Science at the Frontier, on 16 November, based on more than 580,000 messages from 1,819 scientists and a survey of 251 of them.

OpenAI announced the report on X, saying scientists reported nearly doubling their productivity and that it used privacy-preserving methods to see how they use ChatGPT across their work.

What the study measured

The team looked at messages sent between 20 August and 22 September 2026 by scientists in biology, chemistry, mathematics, computer science, physics, engineering and earth science. OpenAI used automated classifiers to sort each message by research activity, such as reviewing literature, writing code or analysing data. Its researchers saw the categories, not the messages or ChatGPT’s replies. By default, OpenAI doesn’t train its models on data from the program.

The survey of 251 participants covered how much time AI saves, what they expect it to do for discovery, and what would still slow research down.

What scientists use it for

Four activities make up 85% of the average scientist’s messages: understanding unfamiliar concepts, drafting papers, analysing results and writing research code. Data gathering, peer review and end-to-end autonomous research are far less common, according to the report’s key findings.

Ideas are a big part of it. OpenAI says 68% of the scientists used ChatGPT to develop research ideas. Across those messages, 35% started with the scientist’s own question, 36% started with a scientist’s idea that ChatGPT helped refine, and 23% had ChatGPT suggest the question. During brainstorming, scientists often treat ChatGPT as a source of ideas; when they develop a research question, they usually bring the first idea themselves.

Bars: scientist's own question 35%, idea refined together 36%, ChatGPT-suggested question 23%, other 6%.
Most ideation messages started with a scientist’s own question or idea; ChatGPT suggested the question in 23%.

How much time it saves

The survey asked scientists to estimate how long the same work, at the same quality, would take without AI. On average they said 1.9 times as long. For research coding they put it at nearly three times, and for research ideation at nearly twice.

Stat card: scientists estimate 1.9x as long without AI; research coding nearly 3x, research ideas nearly 2x.
Coding showed the biggest estimated saving: nearly three times as long without AI.

Scientists kept checking the answers. After a request, they followed up within a week to ask for explanations, checks or revisions about once every three messages, and that held in coding and data analysis too. OpenAI reads that as scientists still exercising their own judgment.

The report also asked which limits would remain even with capable AI tools. Limited time or staffing came up for 51.8% of scientists and funding for 43.8%, and 71% picked at least one of the two.

What scientists expect

Of the 250 scientists who answered the question about discovery, 98% expect AI to speed it up in their field, and 22% say a discovery that would take 10 years without AI could happen within three. Among 224 written explanations, OpenAI counted 367 reasons: 52% were about getting research done faster, 23% about finding more promising questions and 14% about pursuing a wider range of questions.

César de la Fuente, who works on antimicrobial research, is quoted on the study page: “Tools like ChatGPT are helping us accelerate scientific discovery. It’s really a playground for a scientist like myself.”

Free access opens up again

OpenAI published a second post the same day, Broadening access to AI for scientists. Applications for the next round of the program open on Monday 23 November. OpenAI will award 50,000 new grants on a first-come, first-served basis, and each one gives the recipient and up to two collaborators free ChatGPT Plus. That supports up to 150,000 researchers, and OpenAI’s goal for the program as a whole is 200,000, double the target it set when the pilot launched in July.

Checklist: opens Monday 23 November, 50,000 grants, you plus two collaborators, ChatGPT Plus.
One grant covers a researcher and up to two collaborators, and applications open on 23 November.

The pilot gave researchers ChatGPT Pro. OpenAI says its review of pilot usage showed most participants would have stayed within Plus limits, which is why new grants use Plus, and that it’s moving away from lottery-based selection. About 20,000 researchers in 33 countries have joined since July. If you’re already on the waitlist and qualify, OpenAI will email you before applications open with instructions for activating your account.

Universities can also use ChatGPT Edu, and biology labs at US universities can apply for the Rosalind Academic Plan, which gives eligible labs up to 30 GPT-Rosalind seats at no cost for up to 12 months. GPT-Rosalind access needs approval through OpenAI’s trusted-access process.

Applications open at chatgpt.com/academic-researchers/apply.

Sources

Disclosure: I’m an OpenAI, Firecrawl and Devin ambassador.


Discover more from Sulat AI & Tech News

Subscribe to get the latest posts sent to your email.