We’ve talked here at Mind Matters News about model collapse; AI’s trained perpetually on their own data start eating their own tails. Uniqueness falls away and everything becomes bland. As more and more writing is AI-generated, that is becoming a bigger problem.

According to Alex Williams at CACM (the Association for Computing Machinery blog), we may not notice model collapse when it is happening:

People tend to imagine model collapse as some dramatic cliff where a model suddenly starts producing gibberish. The reality is more subtle and, honestly, more dangerous because of that subtlety. What you get is a slow erosion of variance. Outputs become more generic. Edges get sanded down. The model starts producing text that reads fine on a surface level but carries less information per sentence.

Think of it like making a photocopy of a photocopy. The first few generations look almost identical to the original. But by the tenth copy, the image is washed out. Everything is legible, technically; it’s just lost the detail that made it useful.

In language, that loss shows up as homogenization. The model reaches for the same sentence structures, the same hedging phrases, the same predictable cadences. If you’ve ever read a block of text and thought “something about this feels AI-generated” without being able to point to a specific error, you’ve already felt what an AI-native web looks like from the outside post model collapse.

“Model Collapse Is Already Happening, We Just Pretend It Isn’t,” Mar 25 2026

As more and more systems start using AI, perhaps we’ve all noticed that quality. So much writing today sounds like a report from the government about a topic the bureaucrats don’t really want to say anything very specific about. But they don’t want to offend anyone either.

If you get the sense that something you are reading is AI sludge, here are nine hints from “Complete AI Training, beginning with “AI text often reads neutral, overly tidy, and thin on specifics-facts wobble, examples feel generic, endings stop short.”

Nine tells are offered, including “AI often mirrors the expected format too perfectly: a Medium-style essay on LinkedIn, a press-release tone in a casual blog, a five-paragraph structure for anything academic. If the packaging feels generic for the channel, raise an eyebrow.”

You might also try AI writing detectors: Grammarly, GPTZero, and CopyLeaks are all free, at least for typical home use.

AI detectors are not always reliable, of course. But we may as well try them rather than just guess blindly. After all, some entirely human writers are simply unimaginative and it shows. And some AI programs were trained on quality human writing and that might show too. It’s somewhat like getting a second or third opinion before committing oneself.