38% of new arXiv papers now flag as AI-written, up from 0.4% pre-LLM baseline
GenerativeFart · reddit · 2026-11-15
Unslop.run scored the full text of 13,500 arXiv papers from 2021-2026 across ten fields and found 38% of papers sampled in Aug-Oct 2026 read as machine-written — twice the flag rate of early 2025. The chart shows a pre-LLM baseline of just 0.4% before ChatGPT's launch, followed by a steady climb to the 35-38% range by mid-2026. Note this reflects style-detector flagging rather than confirmed AI authorship, but the trend shows how fast AI writing has permeated academic output.
More from AGI Musings
- Researcher now runs AI detector Pangram on papers before deciding how carefully to read — TuhinChakr · 2026-11-15
- Founder: building a bank run by agents, serving agents, defended by agents — neal_lathia · 2026-11-15
- LessWrong essay argues forecasting is way overrated and should stop being funded — JacquesThibs · 2026-11-15
- AI researcher clashes with philosophy fans: frontier AI problems dwarf Hegel — fkasummer · 2026-11-15
- 426 games slated for Steam in 3 days as AI creation floods the market — jhnnassky · 2026-11-15
- Ethan Mollick ranks AI slop replies by annoyingness, from copy-paste to fake curiosity — emollick · 2026-11-15