OpenAI is paying Cloudflare for signals that tell its crawler when a web page has changed, according to Search Engine Journal’s account of the third-quarter earnings call, where CEO Matthew Prince described the deal. The disclosure matters because it puts a price on crawl efficiency, a cost that publishers have so far absorbed without compensation.
The quarter brought Cloudflare’s initial revenue from “Act 4” products, Prince said. Act 4 is the product group aimed at creating a different commercial model for the web. He gave no Act 4 revenue figure, and the results release, which reported $782.5 million in total revenue, up 39% year over year, does not break one out.
The OpenAI arrangement works like this: Cloudflare tells OpenAI when a page has been updated, such as a restaurant changing its prices, so the crawler does not need to keep re-fetching unchanged pages. “OpenAI is paying for that,” he said. He predicts other AI companies will follow. Cloudflare first made the pilot public on July 8, using signals from participating websites such as content freshness and page changes.
Prince did not say what OpenAI pays or when payments began. He also did not address whether the websites in the pilot receive any share. That gap is the central open question for publishers, because their pages supply the signals being sold.
On agent traffic, Prince split customers into two camps. Some customers welcome the maximum number of agents, while “Others see them as a bane to their business and want to keep them away,” he said. Cloudflare helps the first group attract agents and handle the load, and it blocks agents for the second group without charging for that traffic. He described the same split on the August call.
He framed the first camp’s goal as “sort of the new version of SEO.” The framing deserves scrutiny. According to Search Engine Journal, the CEO named no Cloudflare service that optimizes a site for agent traffic. The tools he pointed to are about access and efficiency, not ranking or selection.
Prince’s cost argument, first made on the August call, is easy to follow. An agent choosing a dinner spot might pull menus from 1,000 restaurant sites, and the user visits one. The other 999 sites bear serving costs and see nothing back. A person might view two or three pages, so agent volume dwarfs human browsing. In August he projected that bot and agent volume could reach up to 1,000 times human volume inside five years. This time he expects it sooner.
Payment ideas went further. Prince anticipates micropayments of some kind for agent visits, and he suggested that an agent explaining its purpose could earn a lower price. His example was an agent telling a travel site it is planning a Hawaii trip for a family of four, which could open advertising opportunities for the site. Looking ahead 10 years, he said, “I would imagine that this becomes the largest part of Cloudflare’s business.”
Cloudflare moved two payment products into beta on Sept. 30. With Monetization Gateway, a seller bills an AI agent for each request. Pay Per Use compensates publishers that take up an AI firm’s proposal tied to reported uses. Search Engine Journal noted that Cloudflare’s AI Crawl Control, available on every plan, lets site owners allow or deny individual AI crawlers, while Pay Per Crawl remains in private beta.
For search teams, the practical step is to decide now which AI crawlers to admit, using the controls that already exist. Log-file data showing agent requests against actual referrals will tell you which side of Prince’s split your site belongs on, and it will be the evidence you need if per-visit pricing arrives.
Search Engine Journal, Matt G. Southern, November 8, 2026.
