Comment by SwiftyBug

I assume this is just some form of hallucination. I don't think Google would train its LLM with details of its own heuristics.