Startup Field Guide by Unusual Ventures: The Product Market Fit Podcast · Unusual Ventures

Perplexity CTO Denis Yarats on AI-powered search

April 21, 2024·35 min·3 clips
Perplexity's traffic didn't drop after ChatGPT—it increased, leading them to abandon months of work overnight.
Sandhya Hegde introduces Perplexity as an AI-powered search engine and frames the interview around how the company found product-market fit. Dennis Yaris talks through how he met Arvind and how the founding team came together. The conversation then shifts to the early prototype, built in just two days as a very simple website. The team did not expect much from it. Instead, usage went up rather than down. That created a hard call. They stopped working on text-to-SQL, set aside months of infrastructure work, and went all in on general search. The episode uses that pivot as the clearest example of how market pull can steer a company’s direction. Sandhya keeps pressing on what the team learned from the users who showed up first. Dennis says Perplexity is not trying to win all of web search. He describes the early and current audience as people with a knowledge-worker profile. Those users search to get specific work done, including research, academic work, market work, and financial questions. The conversation draws a line between those queries and recreational or purely navigational searches. The first group is more tightly tied to decisions. Dennis argues that decision-linked searches are more monetizable because they sit closer to real economic value. He points out that Google makes most of its money from a relatively small slice of its user base. Perplexity, he says, overlaps with that valuable slice more than it overlaps with the users who generate only broad free traffic. That defines both the opportunity and the ceiling. He is clear that Perplexity is unlikely to become as large as Google, and that is not the aim. The aim is to build something useful for the people who need it. The interview also touches on the tension between search and generative AI, including the idea that some hallucination can sometimes serve a product purpose. Throughout, the conversation stays grounded in practical lessons from the company’s early choices. It keeps returning to the same pattern: build, watch real usage, identify where the value is, and narrow the product around what customers actually reveal.

As heard by us

A practical look at how early usage pushed Perplexity from text-to-SQL toward general search.

Perplexity's story here is really about a small team following usage instead of sticking to the original plan. They built a basic website in two days, saw traffic keep climbing, and then dropped text-to-SQL to focus on general search.

Read the full review in PlayNext →

Why you'd press play

You want the story behind Perplexity's pivot from a two-day prototype to general search.

Read the full recommendation in PlayNext →
Listen to the show on