Perplexity AI Goes Public — A Search Engine That Answers, with Sources

On 7 December 2022 the San Francisco startup Perplexity AI released its product, "Perplexity Ask", as a public beta. The four founders were Aravind Srinivas (CEO, ex-OpenAI and Google research), Denis Yarats (CTO, ex-Meta AI), Johnny Ho and Andy Konwinski.

ChatGPT had been released exactly a week earlier, on 30 November 2022. Perplexity therefore presented its alternative — "LLM × web search × citations" — before the world had quite finished noticing that ChatGPT was going to be transformative.

Design — A Grounded LLM

The strongest immediate criticism of ChatGPT was hallucination: the model wrote fluently but often invented facts. Stale training data, ambiguous query language, and the model's fundamental "pick the most plausible word" behaviour all blurred the line between writing and fiction.

Perplexity's answer was to ground the LLM in retrieved web content. The pipeline:

1. The LLM interprets the user's query and reformulates it for a search engine.
2. The system retrieves a handful of relevant pages from Bing and from Perplexity's own crawler.
3. The full text of those pages is injected into the LLM's context.
4. The model writes the answer, attaching numbered footnotes to each sentence indicating which source it drew from.
5. The user can follow any footnote to the original web page and verify.

This pattern — now known across the industry as retrieval-augmented generation, or RAG — was first deployed at scale in a consumer search product by Perplexity.

Versus ChatGPT

Looked at on a timeline, Perplexity's positioning is almost startlingly direct.

August 2022: Perplexity AI is founded. Seed round of about US$3 million from Elad Gil, Nat Friedman and others. 30 November 2022: OpenAI releases ChatGPT and the world begins to notice. 7 December 2022: Perplexity Ask launches. It is not a panicked response to ChatGPT — development had been under way for months — but the timing places it exactly a week behind.

ChatGPT is a conversational LLM; Perplexity is an answer engine. The surfaces look similar; the design philosophies are opposite.

• ChatGPT: generate from the trained model's internal knowledge. No citations. No real-time information. Task scope from chit-chat to writing to coding.
• Perplexity: for every query, run a fresh web search and inject the results into the LLM context. Citations are mandatory. Task scope narrowed to "answer the question".

In effect, Perplexity pre-emptively patched ChatGPT's three structural weaknesses — hallucination, stale information, missing sources — with a different architecture.

Rapid Growth — 2023 to 2024

Through 2023 Perplexity's traffic grew rapidly.

February 2023: about 2 million monthly unique visitors. April 2023: Series A of around US$26 million, led by NEA. Valuation US$150 million. July 2023: mobile app shipped; Series B of about US$74 million. Valuation US$520 million. January 2024: Series B extension of another US$74 million, with NVIDIA and Jeff Bezos joining individually. April 2024: Series B2 of about US$62.8 million. Valuation passes US$1 billion — unicorn status. June 2024: WSJ reports Perplexity is negotiating a round at a US$3 billion valuation. December 2024: Series E closes around US$500 million at a valuation of about US$9 billion.

Feature releases ran in parallel: Comet, Pro Search, Spaces, Pages, Shopping Hub. The launch of Comet, an AI-driven browser, in 2024 signalled a direction beyond the search box, toward AI handling browsing as a whole.

2025–2026 — A Valuation of US$21 Billion

May 2025: about 30 million monthly active users; 780 million queries that single month. September 2025: Series E-5 raises US$200 million at a US$20 billion valuation. Early 2026: Series E-6 closes at a valuation of about US$21.2 billion. Monthly active users around 45 million. Annual recurring revenue (ARR) around US$200 million.

From a December 2022 launch to early 2026 — just over three years — that is roughly a 7,000-fold rise in valuation and orders of magnitude growth in users. Perplexity has become the third pole of the US AI-search field, alongside OpenAI's ChatGPT Search and Google's AI Overviews.

The Competitive Picture

As of May 2026, the AI-search landscape divides roughly as follows.

• Google AI Overviews: LLM-generated summaries grafted onto the top of the existing search results page. Monthly active users in the billions. But the reduction in click-through to source sites has become a serious concern.
• ChatGPT Search: launched GA in October 2024, opened to free users through 2025. Hundreds of millions of MAU, though not search-specialised.
• Perplexity: a stand-alone search product. Differentiates on explicit citation and a thoroughly "answer-first" UX. About 45 million MAU.
• Microsoft Copilot / Bing: a GPT-4-class layer over the Bing index, bundled with Microsoft's existing search advertising.

All four converge on the same basic structure: LLM + real-time search + citations. The design Perplexity proposed in December 2022 has become the industry default.

What Changed

What Perplexity changed was the twenty-five-year assumption that search is a list of ten blue links.

From Google in 1998 through to 2022, the standard search UX was: type a query, receive ten ranked links, click. Perplexity inverted that into: type a query, receive a written answer, footnoted with sources at the end, and only those who want to verify open the original.

The inversion has begun to reshape the economics of the wider search market. The traditional model — sites earn traffic by ranking high — is being replaced by a worrying alternative in which sites are cited but the user never clicks. The October 2024 cease-and-desist letter from The New York Times to Perplexity, and the lawsuit that followed in 2025, are emblematic of that structural problem.

What Perplexity built alone is not finished. The combination of LLM and search is rewriting the standard shape of search itself while still carrying unresolved questions of citation reliability, copyright, and the economics of the sites it draws from.