Perplexity
The startup trying to replace Google by letting AI do the searching
Perplexity took the universally despised experience of Googling something and getting 10 SEO-optimized blog posts and said: what if we just... answered the question? Here's how they're building an AI-native search engine.
$9B+
Valuation
250M+
Queries/month
$20/mo
Pro price
Finding the Real Problem
Google search has gotten worse (you've noticed — everyone's noticed). Pages of ads, SEO-gamed content, and sponsored results before you get to anything useful. Perplexity's insight was that people don't want links, they want answers. They built a search engine that reads the internet for you and gives you a synthesized, cited answer.
Trust Through Citations
The killer feature isn't the AI-generated answer — it's the footnotes. Every claim is linked to its source. Users can verify anything. This solves the hallucination trust problem in a really elegant way: "Don't trust us, trust the sources. And here they are." It's a design pattern that more AI products should steal.
The Conversation Model for Search
Traditional search is one-shot: you type a query, get results, done. Perplexity adds follow-up questions. After getting your initial answer, you can drill deeper, ask for clarifications, or explore tangents. This conversational approach means users spend more time on Perplexity and get more value per session.
The Business Model Challenge
Competing with Google on search is, to put it mildly, ambitious. Perplexity's bet is that $20/month subscriptions for power users (more queries, better models, file upload) can build a business while they figure out the ad-supported model. They've also launched publisher partnerships to share revenue with content creators whose work the AI cites.
Lessons for Product Managers
- →Sometimes the biggest opportunity is the thing everyone complains about but thinks is unfixable. (Google search quality, in this case.)
- →Citations and source attribution aren't just nice-to-have — they're a trust mechanism that makes AI outputs fundamentally more useful.
- →The conversational follow-up pattern is powerful. Let users dig deeper instead of starting over.
- →Taking on a giant is possible if you're solving a real pain point that the giant has gotten complacent about.
- →Think about your content supply chain. If your AI uses others' content, figure out how to credit and compensate them.