To Trust or Not to Trust: Authors' Response to AI-based Reviews

In three linesStudy of 56 authors from 40 papers: 83.9% find AI reviews useful, 80.4% report AI identifies issues missed by humans, 82.1% incorporate AI feedback into final version. However, authors trust AI less than humans (51.8% report minor inaccuracies, 16.1% report serious errors). 96.4% would accept AI as internal review tool before submission.

Summary generated by Claude — human-verified