Perplexity AI is an “answer engine”: a search tool that reads a question typed in plain language, pulls up-to-date information from the live web, and writes a short, direct answer with numbered citations to its sources — instead of handing back a page of blue links to click through. Launched in December 2022 by a small team of former OpenAI, Meta AI, and Quora engineers, Perplexity AI has grown into one of the most-used challengers to traditional web search.
What it is
Type a question into Perplexity the way you would ask a knowledgeable friend — “why did this company’s stock drop today” or “what’s the difference between two visa types” — and it returns a paragraph-length answer with small numbered footnotes. Click a footnote and it jumps to the exact web page the claim came from. Follow-up questions stay in the same conversation, so you can narrow or redirect without starting over, much like a chat app.
The company also ships a Chromium-based web browser called Comet, built around the same search-and-answer engine, and mobile apps for iOS and Android.
How it works
Perplexity is built on a technique called retrieval-augmented generation, or RAG. Instead of relying only on what a language model memorized during training — which goes stale and can be wrong — the system first runs a live web search for the question, retrieves a handful of relevant pages, and then feeds those pages to the language model along with the original question. The model’s job shifts from “recall an answer” to “summarize what these specific documents say,” which is why it can cite a source for nearly every sentence.
Perplexity doesn’t rely on one in-house model. Depending on the plan and query, it can route a question to different underlying models — including ones from OpenAI, Anthropic, and Google — plus a proprietary model of its own called Sonar, tuned specifically for fast, web-grounded search. Paid users can also pick a preferred model manually.
How it differs from Google and from a chatbot
A conventional search engine returns links; the reader still has to open pages and assemble an answer. A general-purpose chatbot writes fluent answers but, without a live search step, can state outdated facts confidently and rarely shows its sources. Perplexity sits between the two: it searches like a search engine but answers like a chatbot, with citations attached so a reader can verify the claim rather than take it on faith.
That citation model is also the source of Perplexity’s biggest controversy: several publishers, including the New York Times, Dow Jones, and the BBC, have accused the company of scraping their sites for answers — sometimes bypassing the standard robots.txt opt-out signal — without a licensing agreement, and of occasionally misattributing quotes to articles that never contained them. Perplexity disputes the scraping claims and has since signed revenue-sharing deals with a number of publishers. Readers should treat any AI-generated answer, including Perplexity’s, as a starting point to verify against the cited source, not a final word.
Getting started, and what it costs
Anyone can try Perplexity for free with a web browser or the mobile app, no account required for basic searches. The free tier covers unlimited basic search but caps the more thorough “Pro Search” queries to a handful per day and doesn’t unlock model choice or file uploads.
Paid tiers add capacity and features: as of July 2026, per Perplexity’s own pricing page, Perplexity Pro costs $20/month (or $200 billed annually), and a higher-capacity Max tier costs $200/month for professionals who query it constantly. A discounted Education plan is available for verified students. Because subscription prices change, check the pricing page for the current numbers before committing.
FAQ
Is Perplexity free? Yes — basic search is free indefinitely; paid plans exist for heavier or more advanced use.
Does Perplexity replace Google? Not entirely. It’s strong for research questions with a clear answer; for browsing, shopping, maps, or highly current breaking news, a traditional search engine or dedicated app may still serve better.
Can Perplexity’s answers be wrong? Yes. It reduces — but does not eliminate — the risk of a language model inventing facts, since answers are grounded in retrieved pages. Always check the linked source before relying on a claim.
What is Sonar? Sonar is Perplexity’s own language model, built for its search product and optimized for speed and citation accuracy rather than general-purpose writing.
Sources: Perplexity AI — Wikipedia; Perplexity pricing.