Substack’s new tool tells you who’s been writing their newsletters with AI
TechCrunch Sarah Perez ● Covered by 3 sources
Substack now lets readers scan any post to see how much of it was written by AI versus a human. It's using detection software called Pangram, built right into the app.
Based on reporting by TechCrunch, Sarah Perez — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Substack just gave its readers X-ray vision. The newsletter platform announced this week that it's integrating with Pangram, an AI writing detection tool, so anyone browsing the app can scan a post, comment, or reply and get an estimate of how much was written by a machine versus a person. Anything over 100 characters qualifies.
The timing is interesting given Substack's whole pitch has been built around trust between writers and readers, the idea that you're subscribing to a specific human voice, not a content mill. Short term, this could sting. A lot of newsletters on the platform almost certainly lean on AI more than their authors let on, and suddenly making that visible risks embarrassing people and chipping away at the sense that Substack is a haven from algorithmic slop.
But CEO Chris Best is betting the opposite happens over time. In a chat with Pangram founder Max Spero, Best framed it as reinforcing what Substack has always promised writers: the platform handles the plumbing, but you still have to bring something worth reading. He called the detection feature "good use of AI" specifically because it protects that division of labor instead of letting software quietly take over the part that's supposed to be irreplaceable.
Importantly, Substack isn't building a ban hammer here. Writers get an optional "AI Author's note" to disclose their process voluntarily, and publishers can run Pangram on their own drafts before hitting publish, or dispute a scan they think misfired. The company is positioning it less as enforcement and more as a nudge toward writers explaining how the sausage gets made.
This also puts Substack alongside a growing club of platforms slapping labels on AI output, following social networks tagging synthetic photos and video, and streaming services now flagging or throttling AI-generated music. The pattern is becoming pretty clear: as generative tools get good enough to hide in plain sight, platforms are deciding that disclosure, not prohibition, is the more sustainable answer.
My take — AI-written commentary, not fact-checked reporting
I like this more than most AI-labeling schemes because it doesn't pretend AI-assisted writing is inherently bad, it just refuses to let people fake full authorship for free. Substack's real product has always been the writer's judgment, not their typing speed, and a detection tool that nudges people toward honesty about their process protects that far better than a blanket ban ever could. Every platform sitting on synthetic content is going to face this same reckoning eventually; Substack just got there first.
Read more about this at: TechCrunch