Merton101 · Case 6 · Science and publishing
AAAI gave every paper an AI review, 22,977 in under a day
Reported 15 Apr 2026 · entered the ledger 25 Sep 2026 · last checked 25 Sep 2026 · Side effect
Every main-track submission at AAAI-26 received one identified AI review; the system generated reviews for all 22,977 full-review papers in less than a day; surveyed authors preferred them on technical accuracy.
The institution
AAAI-26 conference
institution
The mechanism
Arms race
externality scale 3 of 5, fast
Adaptation
detect
the institution can tell machine from human, and stops there
The case
- The assumption that broke
- Reading is done by a person, so review capacity limits volume.
- The first-order effect
- The reader automated; authors preferred the machine's review on technical accuracy.
- Who pays
- reviewers whose judgment is now benchmarked against a model · group: Volunteers and reviewers
- Scale and speed
- 3 of 5 · fast
Evidence
-
arXiv ↗
15 Apr 2026
generate reviews for all 22,977 full-review papers in less than a day
The second reader
| Claim | Verdict | Note |
|---|---|---|
| [number] reviews generated for all 22,977 full-review papers in less than a day | supported | Abstract states: "The system combined frontier models, tool use, and safeguards in a multi-stage process to generate reviews for all 22,977 full-review papers in less than a day." |
| [fact] participants preferred AI reviews to human reviews on key dimensions such as technical accuracy | supported | Abstract states: "participants not only found AI reviews useful, but actually preferred them to human reviews on key dimensions such as technical accuracy and research suggestions." |