Google Gemini Didn’t Kill Perplexity — But It’s Eating Its Lunch
The headline hitting Hacker News this morning is “Google Gemini Killed Perplexity AI.” It’s the kind of title that gets clicks and misses the point in equal measure. No single product “killed” Perplexity. But the market share numbers tell a story that Perplexity’s investors are probably reading with one hand over one eye.
What the data actually says.
Perplexity’s share of global AI web traffic dropped from 2.0% in March to roughly 1.3% in May, according to Similarweb data compiled by fatjoe. That’s a 35% decline in two months. Meanwhile, Google Gemini grew from 25.5% to 27.3%. Claude jumped from around 6% to 8.9%. Even Microsoft Copilot doubled, from 1.1% to 2.0% — which means Copilot now has a larger traffic share than Perplexity.
The only major player losing share alongside Perplexity is ChatGPT, which slipped from 56.7% to 52.7%. But losing share when you’re at 56% is different from losing share when you’re at 2%. OpenAI can absorb a few points. Perplexity can’t.
The pivot story is real, but it’s also a confession.
Perplexity’s response to all this isn’t to fight harder for consumer search traffic. They’re pivoting hard to enterprise agents. Computer — their agentic product — is now integrated directly inside Microsoft 365 (Word, Excel, PowerPoint, Outlook, Teams). They launched Personal Computer, an always-on AI agent running on a dedicated Mac mini. They rebuilt their API as a full-stack agent platform with Agent, Search, Embeddings, and a forthcoming Sandbox API.
This is smart. It’s also an admission that they can’t win the consumer chatbot war.
The numbers back the pivot: Perplexity is doing $450-500M in annualized revenue at a $22.6B valuation. That’s about 45-50x ARR. High, but down from the 100x+ multiple they commanded in September 2025. Revenue per user is the metric they’re betting on, not raw traffic share, and from that angle, a company generating half a billion dollars from a fraction of a percent of AI web traffic is genuinely punching above its weight.
The counterarguments — because I’m not going to pretend they don’t exist.
First: Similarweb data is noisy. AI platforms route traffic through apps, APIs, and distributed infrastructure in ways that web-only metrics miss. Perplexity’s mobile app and API usage could be undercounted.
Second: Enterprise revenue is stickier than consumer traffic. If Perplexity nails the Microsoft 365 integration, they could build a durable business that doesn’t need to win the consumer popularity contest. The $750 million Microsoft Azure compute commitment gives them a powerful ally.
Third: Being small means you can move fast. Perplexity doesn’t have Google’s bureaucracy or OpenAI’s safety theater to navigate. Their Personal Computer product — a dedicated Mac mini running an always-on agent — is the kind of weird, ambitious bet a bigger company would kill in committee.
Why I’m still bearish.
The counter-counterargument: every major competitor has a better-funded agentic product in development. Google has Gemini agents. OpenAI has Operator. Anthropic has Claude Computer Use. All three have orders of magnitude more users, more data, and more compute. Perplexity’s agent pivot buys them time, but it doesn’t change the math on who has the resources to win the agent platform war.
The other risk nobody’s talking about: Perplexity’s brand is still “the search engine that cites sources.” If they become “the enterprise AI agent company,” they lose the identity that got them to $22.6B in the first place. Brand pivots at this stage are possible (see: Slack, which started as a gaming company), but they’re rare and painful.
The bottom line.
Google Gemini didn’t kill Perplexity AI. Perplexity is still alive, still generating real revenue, still making interesting bets. But the consumer AI market is a zero-sum game right now, and Perplexity is losing. The enterprise pivot might work. It might even be the right call. But they’re running a marathon on a track where the other runners have longer legs, deeper pockets, and a head start.
I’d love to be wrong. I just don’t have the data to bet that way.
Sources: fatjoe — Perplexity AI Stats July 2026, Similarweb, HN Discussion