Cloudflare introduced Pay Per Use in beta on September 30, proposing that publishers earn money when an AI buyer uses their content downstream. The payable event can therefore occur after a page has been fetched. For publishers, the central question is how the system connects that later use to a credible payment record. Cloudflare’s announcement
The product distinguishes a crawl from the value obtained through a crawl. A page might be retrieved once and then contribute to many answers or recommendations. Charging only for access would count the retrieval; a downstream-use arrangement can price a different event. Cloudflare’s examples illustrate that distinction, rather than demonstrating that participating publishers have already achieved a particular revenue result.
The announced process begins with a buyer defining a paid use and offering a price. A publisher chooses whether to accept the arrangement and can stop participating. The buyer reports uses, and Cloudflare says it bills the buyer and pays the publisher monthly through a connected payment account. Specific terms, including restrictions on training, remain part of what the parties agree to.
Buyer reporting is therefore a material dependency. Cloudflare describes verified-bot identity and checks that reported uses correspond to enrolled publisher content. Its terms require complete reporting. These mechanisms help identify the buyer and validate a submitted record, but the announcement does not establish automatic detection of every downstream use that a buyer fails to report.
My inference is that identity, accounting and completeness answer different questions. Knowing who retrieved content helps assign responsibility. Matching a reported use to a publisher helps allocate payment. Establishing that all payable uses were reported requires additional confidence in the buyer’s behavior and the arrangement’s terms. A publisher assessing the beta should keep those questions distinct.
The announcement also describes richer usage context, such as topics and products, as future work. Current beta participation should be judged on the current reporting and payment design. Pay Per Use offers a concrete attempt to compensate value beyond the initial fetch; its practical usefulness will depend on adoption, agreed definitions of use and trustworthy reporting, rather than the payment label alone.