Why it’s still open
Nearly every study measures outcomes minutes or days after AI use: an exam, a quiz, a second essay. Whether lost practice compounds over months, reverses when people return to doing the work, or simply shifts effort to other skills is unknown. At population scale, one study finds no effect of ChatGPT on test scores so far.
What would count as an answer
Studies that follow the same people for months with and without AI on a core skill, and measure recovery when AI is removed.
Papers that tackle it
- Kosmyna et al 2025 - Your brain on ChatGPT — The origin of the term — small, unreviewed, and contested.
- Stankovic et al 2026 - Comment on Your brain on ChatGPT — The critique.
- Bastani et al 2025 - Generative AI can harm learning — Practice gains, exam losses.
- Huntington-Klein 2026 - Little impact of ChatGPT on test scores — No population-level effect so far.
- Budzyn et al 2025 - Endoscopist deskilling — Possible deskilling in experts within months.
Methods in play
Who’s closest
Placed by the editor, v1 — corrections welcome.
Related questions
Know a paper that belongs here, or think this is framed wrongly? Suggest a paper or a correction (forms on the Atlas home).