Did a Human Write This?

When Max Spiro, co-founder of the AI detection company Pangram, ran the text of a canceled horror novel called Shy Girl through his software, it came back as 78% AI-generated. The publisher Hachette had pulled the book amid suspicions, and the author denied using AI, suggesting an editor had inserted AI writing into a draft. But the incident captures something larger: we are in the early stages of what Spiro calls a "slop pocalypse," where chatbots make it trivially easy to churn out synthetic text at scale, and the internet is shuddering under the weight.

Why detection matters

Spiro's concern starts with a simple asymmetry. "If I'm reading text on the internet, it takes me one minute or five minutes to compose a tweet," he says. "In that time, a large language model can pump out hundreds of thousands of tweets." The result is a collapse of signal-to-noise ratio. Some of this content is obvious spam, but much of it is not. People use ChatGPT to write LinkedIn posts announcing new jobs, Tinder profiles, and earnest comments. They are not trying to deceive—they are trying to communicate. But the effect is the same: the internet becomes harder to trust.