Scientists surveyed by OpenAI estimated that work taking one hour with AI available would have taken about 1.9 hours without it, at the same quality, according to findings the company published November 16, 2026. The figure is self-reported, and OpenAI cautions that its findings don’t establish causal effects. Source
The company’s Economic Research team analyzed 580,186 scientific messages from 1,819 scientists in its ChatGPT for Academic Researchers pilot, covering August 20 through September 22. OpenAI separately presented survey responses from 251 scientists, collected September 30 through October 7. Source
For the productivity estimate, OpenAI averaged responses across eight research activities, giving each activity equal weight. Each included 249–251 respondents; when someone reported using no AI for an activity, OpenAI counted their time as unchanged. The same-quality comparison was part of what scientists were asked to estimate, not an independently verified assessment of their output. Source
What scientists asked ChatGPT to do
According to OpenAI, participants used ChatGPT to explore ideas, analyze evidence, write code and communicate findings. The company reports that 68% used it to develop research ideas, defined as sending at least three messages about that activity during the study. A prompt count is not a breakthrough count. Source
OpenAI says automated classifiers categorized the messages and the research team did not inspect the underlying messages or ChatGPT’s responses. It also says model training is off by default on the academic-researcher plan. Source
Optimism isn’t an outcome measure
OpenAI says 98% of the 250 scientists who answered its discovery-timeline question expected AI to accelerate discovery in their field. Respondents were asked to imagine a discovery that would take 10 years without AI and estimate how soon it would happen with AI. They were forecasting, not reporting completed discoveries with shorter timelines. Source
OpenAI itself says establishing whether those expectations translate into faster discovery will require more research. Scientists also reported constraints on funding, staffing and experimental capacity. Saving time on code or a draft could matter a great deal; it doesn’t make the rest of the laboratory appear on command. Source
Comments