Perplexity Rebuilding the Search Stack
Perplexity: the $11M/year Cliff Notes for the web growing 4,272%
The strategic point is that Perplexity is trying to turn a thin product layer on top of other companies’ search and models into a full search engine that it controls end to end. It starts with hard, long queries where Google often returns SEO pages and link lists, learns from which answers users click, trust, and refine, then uses that behavioral data to replace one layer at a time, retrieval first, then indexing, ranking, and personalization.
-
This is the same pattern used by many infrastructure challengers. Start with the interface where users feel the pain, collect usage data, then move down stack. Perplexity began with cited answer summaries over Bing and Google results, then expanded into internal file search, its own crawler, and a developer search API backed by a continuously refreshed index.
-
What matters most is not just having a crawler, but deciding what to crawl, how to score quality, and how to rank for complex queries. Exa describes the core search advantage as retrieval quality, meaning search that understands intent instead of just matching keywords, and it starts by indexing a high quality subset of the web rather than everything.
-
If Perplexity can own more of retrieval, it gets three benefits at once. Lower dependency on Google and Bing, lower unit costs than paying for outside search results, and a proprietary feedback loop from consumer searches, enterprise file queries, and API traffic that can train better ranking and personalization for high value knowledge work use cases.
This heads toward a market where answer engines split into two camps. Thin wrappers that rent search, and full stack players that own enough of crawling, indexing, and ranking to shape results themselves. Perplexity’s path is to become the second kind, then push that engine into consumer search, enterprise knowledge search, and developer APIs from the same underlying retrieval system.