Perplexity

What it actually does

Perplexity answers questions by searching the live web and writing a summary with citations attached. You get focus modes to narrow a search to academic papers, YouTube, Reddit or specific domains, file upload, and image understanding. The interface is clean and ad-free.

The citation model is the right idea, and it is why researchers reach for this over a plain chatbot. Inline sources mean you can check the answer instead of trusting it, and research-heavy work is the standout use case in reviewer commentary. A Pro subscription also gives access to multiple frontier models, GPT and Claude among them, under one bill. It is the strongest-rated tool in this batch at 4.4 out of 5 across 354 G2 reviews, 70% of them five-star and only 1% at one star.

What it costs

Free gives unlimited quick searches and 5 Pro searches a day. Pro is $20 a month. The free tier is more usable than most: quick search covers ordinary lookups, and 5 Pro searches a day is enough to test whether the deeper mode earns its keep.

The catch on the paid plan is what you are actually buying. Pro subscribers complain the value depends on which underlying third-party model gets routed to their query, and that the routing changes without notice. You are paying $20 for access to models Perplexity does not own, arranged by logic it does not publish, so the product you evaluated in March is not guaranteed to be the product you get in September. Low-star Trustpilot reviews cluster around billing and subscription handling, but Trustpilot blocked the page, so we are not publishing a number.

Where it falls down

The citations do not always hold. Users report answers citing sources that do not support the claim, and summaries built from web traces rather than the article itself. The footnote looks rigorous whether or not it is, which is a specific and nasty failure mode: the thing that makes you trust the output is the thing that is unreliable. If the work carries your name, you have to open the sources yourself, and at that point you are doing the research anyway.

The conduct record is the other half. In June and July 2024 Forbes alleged Perplexity had lifted its reporting near-verbatim into Perplexity Pages with minimal attribution, which means users could unknowingly republish plagiarised text. Wired alleged illicit scraping of its site and other Condé Nast publications. TechCrunch summarised it on 2 July 2024 in reporting by Rebecca Bellan: "Perplexity has ignored the Robots Exclusion Protocol to surreptitiously scrape areas of websites that publishers do not want bots to access." Bellan also wrote that "a series of accusations in June suggests the startup's approach borders on being unethical."

Robots.txt circumvention was documented independently by the developer Robb Knight and by Wired. In 2025 Cloudflare publicly accused Perplexity of using undisclosed crawlers to reach blocked sites, and Perplexity's defence is that it operates as a user-driven agent rather than a crawler. The part that matters for a buyer is the gap between statement and observation: the company said it honoured publisher opt-outs while reporters watched it do otherwise.

Who should buy it

Anyone doing exploratory research who treats the answer as a starting point. Students orienting in a new subject, journalists finding leads, and analysts scanning current events all get real value, and the free tier plus 5 daily Pro searches covers a surprising amount of that.

Do not buy it as a general chatbot, because that is not what it is and you will be disappointed. Do not use it as a citation machine for anything published, legal, medical or client-facing without opening every source, since the recurring complaint is precisely that the citation and the claim can come apart. If your organisation has taken a position on publisher scraping, the Wired and Cloudflare accusations belong in that decision.

Our 3.0 is a useful tool held back by unreliable sourcing and a real trust gap.