I can mostly tell when webpages from my search results are trying to bullshit me (and Kagi gives me some nice tools to suppress the bullshitting kind), but with ChatGPT I have no idea.

Yes true if you trust the website then generally that trust can extend to all its content. You are putting your faith into the competence and consistency of a human being which is generally more trustworthy than the hit or miss results of a word predictor