Study Finds Rapid Rise in AI-Generated Text Across Top Law Reviews
An analysis of flagship law review articles shows AI-written content jumped from 3.3% of papers in 2025 to 15.2% in 2026, with one article over half AI-generated.
Julian Nyarko applied Pangram 3.3.2 to 571 law review articles from the T14 flagship journals, covering 2020, 2025 and 2026 publications. The 2020 set, predating GPT-3, showed no AI signal, confirming low false-positive rates. In 2025, 9 of 273 articles (3.3%) contained any AI trace, with 5 exceeding a 5% AI word share.
By 2026, 25 of 165 articles (15.2%) displayed AI content, including nine with more than 5% AI words and one where roughly two-thirds of the text was AI-generated. AI signals clustered in abstracts and main bodies, while footnotes were largely untouched. The analysis cannot pinpoint whether editors, authors, or other contributors employed AI, nor distinguish between drafting, polishing, or editing uses. Nyarko refrains from judging the practice, offering the data as a descriptive baseline for ongoing debates about AI in legal scholarship.
Why it matters
The findings highlight a swift increase in AI assistance in academic legal writing, raising questions about authorship and scholarly standards.
In this story
