I Stopped Trusting ChatGPT's Facts and Started Here Instead

For months I let ChatGPT end arguments with a confident sentence and called it research. Then I cited a stat it invented in front of a client. That was the last time.

Now, before any fact goes into something someone else will read, it runs through Perplexity. The habit is simple: ask the question there, read the cited sources, only then bring the number back to my writing. The link is the point, not the prose.

The trick that made it stick: I open Perplexity in a side tab and paste the specific claim, not the whole topic. 'What's the cancer survival rate for X' gets me three sources. 'Tell me about cancer' gets me a lecture.

Perplexity still hallucinates sometimes, which is why I click at least one source before trusting a number. But it fails toward honesty, showing its work, while a chatbot fails toward fluency, hiding it.

If you publish or decide based on facts, make source-checking a separate step with a tool built for it. Your credibility is the only thing ChatGPT won't regenerate for you.