For researchers
Ask us to review this record
TRV-2026-2992AI chatbots misread eye scans: GPT 5.2 Pro got 32.1% of diagnoses right in a 330-scan study
If you wrote or worked on the study behind this record and we got something wrong or left out something that matters, tell us here. An editor reads every request, checks it against the paper and emails you with what we decided. If we change the record, the change and the reason go in its public change history.
ace