News and analysis on AI in research, journal policies, peer review and academic integrity. Explore our academic AI Detector to review a draft, or read practical research and writing guides.
arXiv limits each submitter to two submissions per calendar month and three active submissions at any time. The limits work separately, and rejected submissions, held papers and co-authored work are treated differently. Here is how they interact and what teams should check before submitting.
Springer Nature and Google DeepMind announced on 26 October a partnership to co-design an AI peer review assistant, with a prototype now being tested. The release gives no access details, launch date or performance results, and says expert human judgement stays central.
AI detector false-positive rates vary widely by tool, threshold, text length, and language background, so no cross-tool number should be trusted outside the benchmark that produced it. This piece reviews sourced 2026 evidence on Turnitin, GPTZero, Originality.ai, and Pangram, explains why STEM/ESL writing triggers false flags, and lays out an evidence-based way to defend legitimate writing.