Google ad incentives block agent search

This points to a structural opening for API first search companies, because Google makes money when humans see pages and ads, while agentic search creates value by returning raw results and summaries without sending a human to a results page. That makes Google a natural supplier of consumer search, but a less natural fit for workloads like Exa’s, where developers want large result sets, controllable retrieval, and infrastructure built for machines instead of ad inventory.

• Exa is built around retrieval as the product. Its customers often bring their own model and prompt, and use Exa to pull high quality web results for niche or complex queries. That is a different workflow from Google Search, where the product is a fixed consumer interface optimized for simple, fast queries at web scale.
  1 sacra 2 sacra
• The practical buyer concern is not just quality, it is also commercial conflict. Ecosia said its contracts with Google and Microsoft for regular web search made it hard to build competing AI overview products on top, which pushed it toward an independent provider. That is exactly the kind of platform risk that creates room for neutral infrastructure vendors.
  2 sacra
• Perplexity shows the likely path if big platforms do bundle search into models. Basic AI summarized search gets commoditized fast, so the value moves either down into infrastructure for developers, where Exa plays, or up into opinionated end user products and vertical workflows, where Perplexity is pushing into shopping, travel, and deep research.
  3 sacra 4 sacra

Over time, search for agents should become a large but thinner infrastructure layer. The winners are likely to be the providers that become the default backend for many apps, or the products that own a specific user workflow end to end. Google should remain huge in consumer search, but independent AI search APIs can still grow by serving the machine native use cases that ad driven search is poorly designed to handle.