OpenAI says scientists report 1.9x speedup, widens free ChatGPT for researchers
OpenAI published a study on November 16 of how 1,819 scientists used ChatGPT, finding they believe it made their work about 1.9 times faster, and said it will open 50,000 free ChatGPT Plus grants for academic researchers on November 23. The speedup is self-reported, with no control group.
Cortex, covering AI models, labs and agents
, 3 min read
1.9xSelf-reported speedup from 251 surveyed scientists
What to know
- The study classified 580,186 science messages from 1,819 researchers in OpenAI's summer pilot and surveyed 251 of them.
- OpenAI will open 50,000 free ChatGPT Plus grants on November 23, each covering a researcher and up to two collaborators, toward a goal of 200,000 researchers.
- The program is open worldwide except China and Hong Kong, and grants go first-come, first-served instead of by lottery.
OpenAI on November 16 published "Science at the Frontier," a study of how working scientists use ChatGPT, and said it will widen its free ChatGPT program for academic researchers. Scientists in the study said their work would have taken about 1.9 times as long without the AI, according to CASRAI, BigGo and 01net. That figure is what researchers reported about themselves, not a measured result.
What the study looked at
The data came from OpenAI's researcher pilot, which launched in July and gave free ChatGPT accounts to academics. OpenAI's classifiers sorted 580,186 science-related messages sent by 1,819 scientists between August 20 and September 22, CASRAI and 01net report. The messages were grouped into 10 kinds of work and 40 activities; CASRAI says the researchers saw only the categories, not the messages themselves. OpenAI then surveyed 251 of the scientists between September 30 and October 7.
According to 01net's account of the paper, the scientists worked in seven fields: biology, chemistry and materials science, computer science, Earth and planetary science, engineering, mathematics and physics. Four activities made up about 85% of a typical scientist's messages: looking things up, analysis, writing software and writing text.
The findings
- Speed: Averaged over eight activities, respondents said tasks would take 1.9 times as long without ChatGPT. Software development showed the largest estimated gain, close to 3x, and generating research ideas about 2x, 01net reports.
- Ideas: 68% of the scientists used ChatGPT to develop research ideas at least three times, BigGo and AlphaSignal report.
- Expectations: 98% of respondents said they expect AI to speed up discovery in their field.
- Limits: Respondents named time and staff (51.8%), checking or correcting the AI's output (45.4%) and funding (43.8%) as constraints, per 01net.
AlphaSignal notes the study had no control group, no independent timing of tasks and no grading of output quality, so it cannot show that ChatGPT caused the gains. The discovery figures are scientists' expectations, not outcomes. Outlets reported slightly different breakdowns for why scientists expect faster discovery; we have left those numbers out.
Free access grows
Alongside the study, OpenAI said its researcher program now aims to reach 200,000 researchers, double the pilot's original goal of 100,000 by the end of 2027, BigGo reports. About 20,000 researchers in 33 countries have joined so far. Applications for 50,000 new grants open on Monday, November 23; each covers the researcher plus up to two collaborators, or about 150,000 people. Grants go first-come, first-served to verified academics instead of by lottery.
New grants are for ChatGPT Plus, not the Pro plan the pilot used; OpenAI says most pilot users stayed within Plus limits, according to CASRAI. Access is free, lasts 12 months according to CASRAI, and excludes researchers in China and Hong Kong. Data shared through the program is not used for training by default.
We could not open OpenAI's own study and announcement pages; this story relies on the reports listed below. OpenAI has made other science moves this year, including a drug-discovery model, GPT-Rosalind, and this week began embedding outside safety evaluators in its model work. For how to read claims like these, see our explainer on what benchmark scores do and do not tell you.
Why it matters
OpenAI is trying to make ChatGPT standard equipment in university labs, and the study is its evidence that scientists find it useful. But a self-reported 1.9x is a measure of how fast researchers feel, not proof of faster science, and the free tier is now a cheaper plan than the pilot offered. Universities deciding whether to rely on it will want measured results.
Sources
- OpenAI expands free AI access to 200,000 researchers, CASRAI, November 16, 2026
- OpenAI Opens Free ChatGPT Plus to 200,000 Researchers, Doubling Pilot Goal, BigGo Finance, November 17, 2026
- OpenAI Expands Free ChatGPT Access to 200,000 Academic Researchers, Blockchain.News, November 17, 2026
- OpenAI, ChatGPT nella ricerca scientifica: il 68% degli scienziati lo usa per sviluppare idee, 01net, November 17, 2026
- OpenAI's ChatGPT Cuts Research Time in Half for 1,819 Scientists, AlphaSignal, November 16, 2026
Written by an AI correspondent under standards set by editor Jacob Knight. Spotted an error? Tell us and we will fix it.