Are younger researchers picking up AI tools faster than their senior colleagues — and what does that mean for science?

This explores whether junior researchers pick up AI tools more readily than senior ones, and what that gap might mean. The short answer from the collection is yes. A Nature Research Intelligence survey of thousands of researchers found that senior researchers used AI less often than early-career researchers, and the gap held for every task measured Where do researchers actually use AI in their work?. The same survey shows where that use goes. Nearly half of researchers use AI often for gathering information and editing papers, while fewer than a quarter use it for peer review. So the adoption gap is mostly about everyday writing and literature work, not judgment-heavy tasks like evaluating other people's research.

The gap matters because of what AI use seems to do to a career. One large-scale analysis found that researchers who use AI publish about three times as many papers and receive nearly five times as many citations Does AI help individual scientists while narrowing scientific focus?. If early-career researchers are the heavier users, the generational gap could turn into a productivity gap. The same study found a collective cost, though. Fields where AI use spreads shrink their range of topics by about 4.6% and their collaboration by 22%, because AI pulls work toward data-rich problems that are already well studied. Faster adoption by younger scientists could therefore speed up their own careers while making the research agenda they inherit narrower.

Seniority isn't the only divide in the collection. In interviews with 25 AI researchers, people at frontier companies engaged seriously with scenarios where AI automates research, while academics often gave the idea little thought Do AI researchers view automating AI research as a severe risk?. Outside academia, firms more exposed to AI replace freelance workers with AI tools faster and more cheaply than less-exposed firms Do firms substitute labor for AI at different rates?. That suggests adoption builds on itself wherever in-house skill has already built up. Taken together, the collection suggests that who adopts AI depends on age, setting and existing capability all at once. Institutions are only starting to write rules for this, as a recent Nature editorial urges Can AI-generated research outpace peer review systems?.

A caveat: the collection has one direct source on the seniority question, and it is a survey of how often people use AI, not how quickly they started. It doesn't tell us whether senior researchers are slower to adopt or simply choose to use AI less, or whether the gap is closing.