The provost who warned students about AI ghostwriting got caught using AI to write that warning.
The Summary
- Dartmouth's provost published nine articles in 2026 about AI ethics—including a Washington Post op-ed condemning students who outsource writing to AI—that appear to be AI-generated
- His denial statement? Also flagged as 100% AI-written by detection software.
- This joins a pattern: the Australian professor, the nonprofit exec, the AI reporter, the Stanford misinformation expert—all caught using AI to write about AI ethics
The Signal
Santiago Schnell published a Washington Post op-ed warning students not to let AI do their writing. Student journalists at The Dartmouth ran his work through Pangram, an AI detection tool. All nine articles he published this year registered as likely AI-generated. When confronted, Schnell sent a statement claiming he only uses AI as an "assistive tool." That statement? Also 100% AI-written, according to the same detector.
The irony is almost too perfect to be real. Except it is real, and it's not isolated. An Australian professor wrote an op-ed about the importance of authentic student writing. AI-generated. A nonprofit executive published a book about truth in the AI age. Included AI-fabricated quotes. An AI beat reporter lost his job for copying AI-invented quotes into his stories.
"People whose job involves writing about the appropriate uses of AI keep getting caught using AI to do it."
These aren't just embarrassing personal failures. They reveal the core tension of the agent economy nobody wants to name: the people setting the rules don't know where to draw their own lines. University administrators preaching "do your own work" to 19-year-olds while their provost pumps out AI essays. Misinformation experts filing legal declarations ghost-written by language models. Beat reporters covering AI who can't tell real quotes from synthetic ones.
This is what happens when tool adoption outpaces cultural norms by 18 months. We're in the gap. The technology works well enough that even people paid to think critically about it reach for the shortcut. But it doesn't work well enough to be invisible, so they get caught.
Key patterns emerging:
- Detection is still possible, at least in academic contexts where people are motivated to look
- The "assistive tool" defense is becoming the standard excuse—and it's paper-thin
- The people setting institutional AI policy are often the least equipped to model responsible use
The Dartmouth case is particularly rich because the university literally coined the term "artificial intelligence" at a 1956 conference. Seventy years later, its leadership can't figure out what counts as their own writing. President Beilock published an Atlantic essay titled "Why I Want More AI at Dartmouth" a month before this broke. Now she's launching a review of her provost's AI use. The message to students: we don't know the rules either, but you'll get in trouble if you break them.
The Implication
This isn't a "don't be a hypocrite" problem. It's a "we haven't defined the job yet" problem. When a provost uses AI to write about AI ethics, is that different from using spell-check? Grammar tools? Research assistants? If the AI drafts and you edit, who wrote it? If you outline and the AI fills in, does that count?
Nobody knows. Not professors, not administrators, not journalists, not researchers. The rules are being written in real time by people who are also trying to use the tools. Watch for institutions to start drawing harder lines not because they've figured out the philosophy, but because the PR risk of getting caught is too high. That's not principle. That's damage control.