IMO the single largest issue was analysts' willingness to believe and report something facially implausible, without checking other data sources.
Which BTW is the problem folks have interacting with LLMs generally (unrelated to surveys). I.e., the problem is logic 🤣
