Perplexity
AI

The story goes: Aravind Srinivas, a PhD from UC Berkeley who had already worked at OpenAI, Google Brain, and DeepMind, built a small LLM tool to help himself research insurance plans. It worked. Not brilliantly, not magically - just well enough that he thought, "this should be a product."

Then ChatGPT launched in November 2022, and the window looked like it might be closing before it had opened. Srinivas and his three co-founders - Denis Yarats (Meta AI Research), Andy Konwinski (one of the founders of Databricks), and Johnny Ho (ex-Quora, ex-Tower Research Capital) - shipped Perplexity to the public on December 7, 2022. One week after ChatGPT. Not six months. Not a year. Seven days.

That tempo - ship first, refine later - has stayed with Perplexity ever since. It is not a company that waits for perfect conditions.

Traditional search engines give you ten blue links and a couple of ads. You click. You read. You cross-reference. You maybe find what you were looking for twenty minutes later. Perplexity skips that loop. It reads the web in real time, synthesizes the relevant information, and hands you the answer - with numbered citations you can actually check.

The company calls itself an "answer engine," a deliberate distinction from "search engine." Where Google's job is to organise information, Perplexity's job is to hand it to you already organised.

This sounds simple. In practice it required threading together real-time web access, large language models capable of careful citation, and a UI simple enough that your grandmother could use it and sharp enough that a researcher would want to. The free tier gets you to the answer. The Pro tier ($20/month) gives you access to multiple frontier models - GPT-4, Claude, Gemini - letting you choose your reasoning engine for any given question.

For tasks that previously meant half an hour of tab-switching, Perplexity can do the job in under a minute. That is a genuine shift in how people work with information, and it shows in the usage numbers: 780 million queries per month as of late 2025.

In February 2026, Perplexity dropped its ad-supported model entirely and went subscription-only. The stated reason: ads create an incentive to bias answers toward advertisers. Whether that framing is principled or a PR move depends on your level of cynicism. Either way, it removed a source of editorial conflict that plagues every other search product.

Students, military veterans, and US government employees can currently access Pro for free - a deal Srinivas announced in early 2026 that generated considerable goodwill and considerable sign-ups.

There is a pattern to how Perplexity operates. When TikTok was hours from being banned in the US in January 2025, Perplexity filed a formal merger proposal for TikTok's American operations. When ChatGPT launched in late 2022, Perplexity shipped a competing product the following week. When the DOJ started eyeing Google Chrome as a potential divestiture target, Perplexity publicly announced it wanted to buy it.

None of these moves required the deals to close. What they required was velocity and visibility. Perplexity has understood, from the beginning, that in AI the war for attention is as important as the war for technology. You ship fast, you say the bold thing, you get the press coverage, you acquire users who would have gone elsewhere.

The numbers suggest the strategy is working. The company had 2 million monthly visitors in March 2023. By early 2026 it had 45 million monthly active users. That's not organic drift. That's deliberate acceleration.

Perplexity's fundraising timeline reads like a speedrun. Seed round closed September 2022. Series B with Jeff Bezos and NVIDIA by January 2024. Unicorn by April 2024. $9B valuation by late 2024. $20B by September 2025. Total raised: over $1.7 billion.

The investor list is its own kind of story. NVIDIA. Jeff Bezos. Tobias Lutke (Shopify's CEO). Andrej Karpathy (formerly of OpenAI and Tesla). Yann LeCun (Meta's chief AI scientist). Ashish Vaswani (one of the original "Attention Is All You Need" authors). And, somewhat unexpectedly, Cristiano Ronaldo in December 2025.

When the inventors of the transformer architecture are writing you checks, it signals something beyond typical venture confidence. In January 2026, Perplexity committed $750 million to Microsoft Azure infrastructure - a sign that compute at scale is now a core part of the strategy, not just an operational detail.

Perplexity's model depends on reading the web. Publishers, predictably, have noticed. The BBC, New York Times, Dow Jones, Reddit, and several Japanese newspaper groups have filed legal complaints against the company, alleging that Perplexity's crawlers bypass robots.txt restrictions and spoof user-agent strings to access content that site owners have asked not to be scraped.

This is not a fringe issue. It sits at the centre of how AI companies can build businesses on top of content they did not produce and do not pay for. Perplexity has, in certain cases, entered into licensing discussions with publishers. But the broader legal landscape remains unresolved, and it represents the single largest structural risk to the company's model.

Srinivas has pushed back on the plagiarism framing, arguing that Perplexity drives traffic to original sources through its citations. Whether that argument holds up in court is a different question.