It seems the author fed a list of documents and points to an AI, asked it to write an argument on it, then the AI hallucinated quotes and tried to badly extrapolate from real data. I think the original point might be true, but it is hard to trust tbh.