If I’m sympathetic to the company at all here, it’s because the tools we have for detecting AI-written text are still pretty bad. OpenAI shut down its own detection tool in June after determining its accuracy rates were too low. (The company has said it is working on other approaches.) But Amazon could still do a better job at removing AI spam, if it would only invest in the customer-service tools that would flag issues like Friedman’s before they turned into viral blog posts. It could also introduce more friction into the account creation process, attempting to verify an author’s identity or at least ask them to disclose when they are using a pseudonym. These and many other signals could help the platform fight spam in a way that cut down on fraud.