LLM-Generated or Human-Written? Comparing Review and Non-Review Papers on ArXiv

Summary

Evidence-based audit of arXiv's own policy decision to ban unpublished CS review papers on the grounds that they are largely LLM-generated. Using two detection methods it confirms LLM content is more prevalent in review papers, but shows that in absolute counts LLM-generated NON-review papers are almost six times more numerous, so the policy misses most of the problem while hitting some subfields hard (Computers & Society faces potential cuts of ~50%). A model for evaluating publishing-platform policy with data rather than assertion.

Contribution

Queued — not yet read.

Method

Queued — not yet read.

Relevance to RISE

Surfaced by the 2026-08-29 literature sweep as a high-priority candidate for the RISE knowledge base (angle: AI in peer review). Verified against: arXiv abstract page opened 2026-08-29 (title, 2 authors, submitted 19 Jan 2026, full abstract read)

Critique / open questions

Queued — not yet assessed.

Key quotes

ArXiv recently prohibited the upload of unpublished review papers to its servers in the Computer Science domain, citing a high prevalence of LLM-generated content in these categories. (abstract)

However, this decision was not accompanied by quantitative evidence. (abstract)