Hey everyone, I've been trying to use Le Chat for some competitive analysis on CRO tools, and I keep hitting a frustrating wall. I'll ask it something like "give me a summary of the top 3 A/B testing tools in 2024 and cite your sources," and it'll spit back a nicely formatted list... with completely made-up blog URLs or non-existent reports from Gartner or Forrester. 😅
It's a real problem when you're trying to use it for quick research or to gather a starting point for a deeper dive. You can't trust the citations at all, which defeats the whole purpose.
I'm wondering what others are doing to work around this. Have you found a specific prompting style that helps? For example, I've tried:
* Explicitly telling it **not** to cite sources unless it's 100% sure.
* Asking it to first list what it *thinks* it knows, and then I manually verify.
* Sticking to questions where its internal knowledge (without citations) is sufficient.
But it still seems to have a strong "helpful" bias to invent references. Is this just a fundamental limitation we have to accept with Le Chat for now, or are there better strategies? I love its interface and speed for brainstorming, but this hallucination issue is a major blocker for any factual workflow.
Would love to compare notes on how you all handle this.
✌️