The American Association for Cancer Research (AACR) used Pangram’s AI text detection tool to analyse 46,500 abstracts, 46,021 methods sections and 29,544 reviewer comments submitted to its 10 journals from 2021 to 2024. AACR staff Daniel Evanko and Michael Di Natale presented the results at the Peer Review Congress in Chicago in September 2025.
Benefits & outcomes
- AACR reports that in 2024 about 23% of abstracts, 10% of methods sections and 4.8% of reviewer comments showed AI-generated text.
- AACR reports that detections in reviewer comments fell by more than half in the fourth quarter of 2023, after its journals banned generative AI in peer review, and then began rising again.
- AACR reports that authors at institutions in English-speaking countries were less than half as likely to use generative AI.
- AACR reports that manuscripts with AI-generated text were more likely to be rejected before review.
- Pesquisa FAPESP, reporting on the analysis, gives 36% of the 7,177 manuscripts submitted between January and June 2025 as flagged by Pangram, with authors declaring AI use for 9% of total submissions.
Notes & quotes
- The authors say the detection tool could not distinguish fully generated text from translation or editing.
- The authors conclude that journal policies have had limited effect on use of generative AI.
- Daniel Evanko said in the same talk that AACR journals had rejected all AI-written Letters to the Editor without consideration until April 2025.
Publisher / company: American Association for Cancer ResearchPangram Labs
Date of case study: 2025
Status: Not stated
Built by: Publisher or platform
Sources:
- Quantifying and assessing the use of generative AI by authors and reviewers in the cancer research field, Peer Review Congress abstract
- Rastros do ChatGPT, Pesquisa FAPESP, 23 October 2025
- Peer Review Congress Chicago, Day 1, Science Integrity Digest, 16 September 2025
- Bahar Mehmani, LinkedIn, 3 September 2025