New AI Detector Is Falsely Accusing Real Writers of Being Robots. Should We Trust It?

Substack just launched a tool that lets readers scan posts to see if they were written by AI or a real person. It uses technology from a company called Pangram, which claims over 99% accuracy. But already, well-known human writers like Freddie deBoer and Taylor Lorenz have been falsely flagged. Even Pangram’s own CEO says the detector should never be the final judge. So why are we letting a machine decide who’s human?

The problem goes deeper than a few false alarms. Critics say this kind of technology creates a culture of suspicion where every writer becomes a suspect and every reader becomes an investigator. One author reportedly rewrote her entire book just to satisfy an AI detector’s score. And now startups are popping up to help people evade the detectors. It’s an arms race that nobody asked for, and honest people are getting caught in the middle.

If you write online, this matters to you. A false AI accusation can damage your reputation, and there’s no easy way to fight back when the accusation comes from a black box that even its makers don’t fully understand. So what do you think, are tools like this helpful or are they doing more harm than good? We’d love to hear your take.