How ChatGPT picks sources — and why you can't trust it blindly

OpenAI's got a search tool bolted onto ChatGPT. When you ask it something, it grabs web results, then feeds those URLs to the model. Sounds straightforward. It isn't.

The real question is what the model does with those URLs. Does it actually read the pages? Does it remember which facts came from which source? Or is it just matching keywords and pretending to be honest about it?

The answer is somewhere in the middle. ChatGPT's search tool is genuinely useful for staying current, but the model can still get things wrong. It can cite a source that doesn't support what it's claiming. It can mix up details from different pages. It's not a search engine — it's a chatbot with a search engine in its pocket.

Why this matters for us: la gente is already using ChatGPT for work, for recipes, for the cousin's side business — and we need to know when it's actually reading the source versus just sounding convincing.