Pangram, an AI text detection company, spent part of this year paying journalist Rod Breslau to publicly shame people on social media for suspected AI use — a partnership the company has since ended, though CEO Max Spero continues calling out alleged AI use based on the company’s scores. Breslau told WIRED he wanted “a harsher approach” because people were “getting off too easy.” His rationale: if you’re a tech CEO or executive building a personal brand, you should write your own posts, and if you don’t, people should make fun of you.
The problem is that Pangram’s technology can’t actually tell the difference between someone who prompted an entire post into existence and someone who used AI to smooth out wording, translate, or edit their own thinking. The score only measures whether AI was likely involved — not how. The author of the original article tested this directly: an English translation of his German original, drafted with AI and edited by hand, scored 28 percent AI. Pangram flagged the final paragraphs as AI-written even though every section was created the same way, with the same amount of human intention.
That false precision has real consequences. Academic institutions reject papers based on percentage scores like these, and many of those papers come from researchers writing in a second language or who are simply better at science than prose. They use AI to express their results more clearly — and Pangram penalizes them for it. The stigma is baked into the business model: if AI use becomes socially acceptable, there’s less reason to pay for detection. Spero now says the company is pivoting to position itself as the go-to authority for originality checks in publishing and education, assuming light AI editing will become increasingly accepted. But the public shaming approach has already done its damage.
The historical parallel is hard to ignore. Alexandre Dumas had assistants like Auguste Maquet write rough drafts that he then refined into classics like The Three Musketeers. An AI detector applied to Dumas would have flagged “Maquet usage” and missed everything that made the books enduring. The quality came from the concept at the start and the polish at the end — a nuance no percentage score can capture.
