Thursday, November 19, 2026 Updated Briefings Saved 中文
AI Daily Radar

AI and the frontier tech around it, explained every morning

Available

Pangram flags 99% of award-winning Peruvian novel as AI-generated

Photo: ulleo / Pixabay

The AI-text detector Pangram classified 99 percent of a prize-winning Peruvian novel as machine-generated, and more than 80 percent of the author’s personal texts as well, according to The Decoder. Mexican publisher Almadía said it conducted its own review of the award and is standing by the author rather than rescinding the prize. The case lands as French publisher Gallimard reportedly plans to begin screening submitted manuscripts with the same tool, putting a commercial detector’s error rate directly in the path of acquisition decisions. The report does not name the author or the novel, and does not include any response from Pangram or an independent test of its accuracy.

Verified and unconfirmed

  • Verified: The Decoder reported that Pangram flagged 99 percent of the novel and over 80 percent of the author’s personal writing as AI-generated, and that Almadía said it reviewed the matter and stood by its award decision.
  • Not yet confirmed: The identity of the author and the novel, Pangram’s own explanation or confidence scores, and any independent replication of the detection results; Gallimard’s screening plan is reported without a start date, scope, or how flags would be used in editorial decisions.

Background

Pangram is a commercial AI-text detector marketed to publishers, educators and law firms; it publishes a false-positive rate of roughly one in 10,000, according to a review citing the vendor’s figures. The dispute comes as French publisher Gallimard plans to start screening manuscripts with the tool, putting a vendor score in the path of literary acquisition decisions.

Why it matters

Pangram’s own publisher-facing material promotes a 1-in-10,000 false positive rate as the reason editorial teams can screen submissions with it, yet in this case the tool scored 99 percent of an award-winning novel as machine-written — a gap authors and editors should treat as a reason to require human review before any manuscript is rejected or an author is publicly accused. The practical concern is sharpest for writers submitting to houses that adopt screening, since Gallimard’s planned rollout means a detector score, not an editor’s reading, could become the first gate a manuscript has to pass.

What to watch

Watch for Gallimard to confirm how and when its manuscript screening begins, and for Pangram to publish a response or revised thresholds for literary fiction.

Further reading

  • [AI Detector : Free AI Checker for ChatGPT, Claude & Gemini Pangram](https://www.pangram.com/)
  • [Pangram AI Detector Review: What the Evidence Shows TextPulse. ai](https://textpulse.ai/blog/pangram-ai-detector-review)
  • Pangram (AI detector) - Wikipedia
  • [AI Detector for Publishers: How Editors Check Pangram](https://www.pangram.com/use-cases/publishing)
Comments

Comments

Comments use a GitHub account and are moderated. Comment rules

More in AI

See all