Perplexity

Perplexity is the one to use when you need to check the answer, not just read it. Every claim carries a link, the search is live rather than remembered, and in our testing its citations were the most consistently accurate of the five — the others cite too, but hallucinate or misattribute sources noticeably more often.

That focus makes it excellent for a specific job: market research, competitive checks, factual questions where being wrong is expensive, and any situation where you have to show a colleague where the number came from. Its Deep Research mode produces genuinely useful sourced reports, and the free tier is generous enough to establish whether the workflow suits you.

It is a poor general assistant, and it does not pretend otherwise. Long creative writing, code generation and extended multi-turn work are all weaker than the four products above. Reviewers who tried to make it their only tool were consistently the unhappiest in our data; those who use it alongside a general assistant are the happiest.

Our guidance: treat Perplexity as a second subscription, not a first. At $20 a month next to ChatGPT or Claude it is a sharp research instrument. As a sole assistant it will frustrate you within a fortnight.