Cloudflare has launched a beta Web Search API that lets AI agents and applications query the live internet instead of relying on stale training data or hallucinated URLs. The API runs through Cloudflare's AI Gateway, meaning all requests are logged, metered, and billed at provider list prices with no markup. Three search providers are available at launch: Ceramic.ai, Exa, and Linkup. The architecture is straightforward: developers call a REST endpoint or use a Workers AI binding, specify a provider and query, and get back search results. Cloudflare acts as the routing and billing layer. All three providers have committed to Zero Data Retention for requests through Cloudflare and to Cloudflare's verified bot crawling standards. The verified bot crawling requirement is the most consequential detail here. Cloudflare already sits in front of a massive share of the web — by some estimates 20% or more of all websites use its CDN. By tying search API access to its own crawling standards, Cloudflare is effectively building a gatekeeping layer that determines which AI search providers can access web content at scale and under what terms. For developers, this is genuinely useful infrastructure. Grounding AI responses in live search results is a real problem, and having a single API that abstracts across multiple providers with unified billing and logging reduces friction. The bring-your-own-key option and no-markup pricing suggest Cloudflare is playing for platform lock-in rather than per-query rent extraction — at least at launch. The three-provider market at launch is thin. Ceramic.ai, Exa, and Linkup are not household names. Google, Bing, and Brave Search are absent. Whether this ecosystem grows or stays a curated garden will determine whether the API becomes a genuine marketplace or a chokepoint. Cloudflare's market power in CDN gives it unusual leverage to shape who participates. The Zero Data Retention commitment is a meaningful privacy signal, though enforcement depends entirely on contractual compliance by the providers. There is no technical mechanism described that would prevent a provider from retaining data outside the Cloudflare pipeline. Trust, not architecture, is doing the privacy work here. This is a classic platform play: provide the plumbing, capture the telemetry, become indispensable. If AI agents increasingly need live web search, the company that routes those queries controls a bottleneck worth far more than the per-query fee.
