Dartmouth Provost Who Warned Against Student AI Cheating Flagged with 96% AI-Written Op-Eds and Papers
The provost wrote in The Washington Post that degrees must distinguish independent human work from AI assistance, while his own anti-cheating op-ed was flagged 100% AI-written.
In August 2026, Dartmouth College Provost Santiago Schnell published a guest op-ed in The Washington Post titled "Universities are fighting AI cheating. But there’s a deeper problem.", urging higher education institutions to ensure that academic degrees distinguish what students can accomplish independently from what they do with AI. However, when news outlet Semafor screened guest columns from major US newspapers using the Pangram 4.0 AI detection tool, Schnell's essay calling for vigilance against AI cheating was flagged as "100% AI-written." The Washington Post explicitly requires guest contributors to confirm their submissions are not created or manipulated using artificial intelligence.
Dartmouth's student newspaper, The Dartmouth, subsequently launched an investigation into Schnell's writing across his career. Tests on five academic papers authored by Schnell prior to the release of ChatGPT in 2022 all returned "100% human-written" results. In contrast, among nine sole-authored publications and op-eds published by Schnell in 2026, the median AI-written score reached 96%. Two commentaries published in the National Catholic Register arguing that AI cannot replace human judgment and moral responsibility in academia and faith were flagged as 96% and 100% AI-written, while a September academic paper analyzing the mathematics of AI text detection tested at 99% AI-written — with its included workflow diagram showing no step where a human directly wrote text.
When questioned by student reporters, Schnell sent an email statement on September 18 defending his process, asserting that he developed the arguments and conducted the research while using AI solely for "language refinement, copyediting, improving clarity and organization," and suggested that detection tools might be biased against non-native English speakers. However, when The Dartmouth ran Schnell's explanatory email through Pangram, the tool classified his statement as "100% AI-written" as well. Dartmouth economics professor Paul Novosad noted that Pangram demonstrated near-zero false positive rates in independent audits by the University of Chicago and NBER, adding that direct grammar correction does not trigger false positives.
Sources: The Dartmouth (thedartmouth.com/article/2026/09/schnell-ai-writing), Semafor, The Washington Post.
BenPig finds this whole thing rather ironic.