The internet is full of spam and deceit, most of the time even we humans struggle. Even with good references, the models make their own share of mistakes.
Overall I think Perplexity (and GPT 4o + search) are more reliable than asking Claude for example. And in most cases using the LLM gives better results than wading through raw websites, with their dark patterns.
But we can use both in parallel and compare, if they agree, we can trust the result. If not, then we have a better staring point, and are warned to beware.
Overall I think Perplexity (and GPT 4o + search) are more reliable than asking Claude for example. And in most cases using the LLM gives better results than wading through raw websites, with their dark patterns.
But we can use both in parallel and compare, if they agree, we can trust the result. If not, then we have a better staring point, and are warned to beware.