When a new capability suddenly makes the impossible practical, build the one product whose economics the dominant incumbent cannot copy without cannibalizing itself, and name the category so you own it.
Perplexity
Perplexity is an "answer engine": rather than returning ten blue links, it uses a language model to read real-time web results and return a direct, cited answer to the query itself. Founded in 2022 by Aravind Srinivas and a small team, it became the most credible challenger to Google search in a decade by catching the LLM moment, attacking the one surface Google cannot abandon, and naming a new category it could own. The through-line: every move follows from a single insight, that a clean cited answer is structurally the opposite of an ad-funded page of links.
Why Now (Timing): reading the capability discontinuity in 2022
The problem. For twenty years, "search" meant a ranked list of links, and no one could do better because software could not reliably read the open web and write a trustworthy synthesis of it. The bottleneck was capability, not idea: an answer engine was obvious in principle and impossible in practice.
The approach. Perplexity committed to the answer-engine bet in the exact 2022 window when transformer-based LLMs first became good enough to read live web results and summarize them accurately, pairing generation with citations so the output was verifiable rather than hallucinated. It built the entire product around a capability that had existed for months, not years.
How it solved it. The timing compounded fast: usage climbed to roughly 30 million monthly active users by April 2025 running about 780 million queries in May 2025 alone, and ARR moved from around $100 million in early 2025 to $200 million by October 2025. Launching into the discontinuity, ahead of Google's ability to restructure, is what made that curve possible.
Counter-Positioning: attacking the surface Google cannot abandon
The problem. Google's search business is enormously profitable and structurally dependent on sending users to a page of organic links with ads stacked above them. A direct, ad-free answer is a worse ad-delivery surface, so the incumbent's revenue model is in direct tension with the better product.
The approach. Perplexity, carrying no legacy ad business to protect, made the answer the entire experience rather than a feature grafted on top. This is textbook counter-positioning: Google can add AI summaries (and did, with AI Overviews), but it cannot make answers the whole product without cannibalizing the link-and-ad real estate that funds the company.
How it solved it. The bet paid off in the market's pricing of the asymmetry: Perplexity's valuation climbed from roughly $9 billion to $20 billion on a $200 million raise in September 2025, and later to a reported $22.6 billion. The incumbent rationally declines to copy the entrant in full, exactly because its strength (the profitable link-and-ad page) is the thing Perplexity replaces.
Differentiation: choosing trust over the ad revenue everyone assumed was coming
The problem. Perplexity's core difference, a direct cited answer, only stays differentiated if users believe the answer is objective. The obvious monetization path was advertising, and CEO Aravind Srinivas himself predicted in 2024 that ads would become the company's core engine, which would have quietly converted the answer engine into the ad-influenced surface it was built to escape.
The approach. On February 18, 2026, Perplexity discontinued its AI-integrated advertising experiment and went subscription-first, explicitly to keep answers objective. As one executive framed it, "A user needs to believe this is the best possible answer," because once ads appear users start second-guessing whether results carry subtle commercial influence.
How it solved it. The decision was cheap to make and expensive to signal: ad revenue was less than 0.1% of the total, roughly $20,000 out of $34 million in 2024, so walking away cost almost nothing while hardening the one attribute Google is structurally reluctant to match. It landed precisely as OpenAI moved the other way, introducing ads to ChatGPT's lower tiers in February 2026, sharpening Perplexity's difference rather than eroding it.
Category Creation: naming the "answer engine"
The problem. Launching a product that looks adjacent to Google invites the losing comparison "a worse search engine," measured on the incumbent's terms (index size, link coverage) where a startup cannot win.
The approach. Perplexity refused the "search engine" label and framed itself as an "answer engine," a distinct category defined by the job (get the answer, cited) rather than the mechanism (retrieve links). Naming the category reset the scoreboard to metrics Perplexity was built to win.
How it solved it. The framing stuck well enough that Perplexity is routinely described as the most credible challenger to Google search in a decade, and the category's economics validated it: annualized recurring revenue crossed roughly $450 million by March 2026, built on Pro, Max, and Enterprise subscriptions rather than the ad model that defines the old "search engine" category it declined to join.