Walled Gardens Make Retrieval APIs Essential
Will Bryk, CEO of Exa, on building search for AI agents
This shift turns search from a general purpose index into a supply problem, and that is exactly why AI native retrieval APIs matter. The best answers are no longer won just by ranking public pages well. They are won by getting access to content that is harder to crawl, keeping indexes fresh, and returning raw results and full text that agents can actually work with. That favors infrastructure companies like Exa over simple SERP wrappers.
-
In practice, fragmentation means the highest value content is increasingly gated by platform rules or commercial deals. OpenAI partnered with Condé Nast in August 2024. Google expanded its Reddit partnership in February 2024. Axel Springer also expanded its Microsoft partnership in April 2024 around AI and premium content distribution.
-
That changes what customers buy from search APIs. One Exa power user runs 5,000 queries a day, pulls 50,000 to 100,000 results, screens them with LLMs, and values Exa most for result depth, full text extraction, and precision on vague queries. The core job is not answering a question once. It is feeding downstream agents and data pipelines every day.
-
Perplexity shows the adjacent model. It built a consumer product on top of Google and Bing SERPs plus multiple LLMs, then monetized with subscriptions, enterprise seats, and ads. Exa sits lower in the stack. It sells the retrieval layer itself, where customers care about raw coverage and control more than the final chat interface.
The market is moving toward a split structure. Consumer products will lock up demand with branded interfaces and content deals, while infrastructure vendors that can aggregate, normalize, and reliably extract from a messier web will become essential picks and shovels for agents. As more software starts searching autonomously, retrieval quality and access rights become the moat.