Substack Integrates Pangram AI Detector as 61% of Writers Report Using AI Tools
Updated
Updated · WIRED · Sep 2
Substack Integrates Pangram AI Detector as 61% of Writers Report Using AI Tools
3 articles · Updated · WIRED · Sep 2
Summary
Late July brought Substack’s integration of Pangram, letting readers scan posts for possible AI use as the platform says it is not anti-AI but wants users to know what they are consuming.
Pangram’s rise reflects growing demand for verification: a 2025 Gotham Ghostwriters survey found 61% of 1,481 working writers use AI tools, while a Stony Brook study of 14,419 self-published novels found nearly 20% had substantial AI-detection scores.
The 24-employee Brooklyn startup has raised $13 million, including $9 million in July, and says its latest model cuts false positives to 0.0041%, though it still misses heavily AI-rewritten human essays 41.37% of the time.
Publishing has become Pangram’s flashpoint after high-profile accusations tied to books and prizes, prompting some publishers and agents to use detection tools while stressing they are only one part of editorial review.
Critics say detectors can wrongly damage careers and may embed bias against some writers, while Pangram argues more transparency and finer-grained detection will make it a broader originality gatekeeper across publishing, education and other fields.