Cloudflare opened a beta on 30 September called Pay Per Use, which charges AI companies when a publisher’s content appears in their product rather than when a crawler fetches the page. It is a different billing event from the one the industry has been arguing about for a year.

The company’s framing of the problem is a sentence most publishers would recognise: “Publishers shouldn’t have to choose between blocking every agent and giving their work away.”

How the mechanism works

There are four steps. An AI company identifies its crawler, states what use it will pay for — Cloudflare’s example is an excerpt appearing in search results — sets a price and connects a payment account. Publishers see the resulting offers in the Cloudflare dashboard under Monetize, and decide whether to accept. The buyer then reports each use as a JSON record carrying a timestamp, a URL and an event ID, through a single API. Cloudflare aggregates the reports, bills the buyer and pays the publisher monthly.

A person reading a news website on a laptop
Publishers review each buyer's offer in the Cloudflare dashboard before accepting. Illustrative photo. Anna Keibalo · pexels · Pexels License

Crawling itself is not charged. Publishers keep separate control over which crawlers may access the site at all, and the programme terms set out what a buyer may do with content it has paid to use, including restrictions on training.

The part that will be argued about

The meter sits on the buyer’s side. An AI company reports its own usage, and Cloudflare says it verifies those reports — without saying in the announcement how. A publisher taking part is agreeing to be paid according to a count it does not produce, for uses that happen inside a product it cannot see.

Large rolls of newsprint paper in a printing plant
Cloudflare aggregates reported uses and pays publishers monthly. Illustrative photo. Mike van Schoonderwalt · pexels · Pexels License

That is not unprecedented — music royalties and ad verification work on similar reported-usage models, with auditing attached — but those systems took decades and litigation to get their auditing rights. This one launches with a vendor’s assurance.

Where it sits against the alternatives

Two other approaches are live. Cloudflare’s own Pay Per Crawl charges at fetch time, and its Monetization Gateway, which moved to closed beta the same day, lets a seller price individual requests from agents and settle in stablecoins. Separately, Google has begun paying a group of about 100 publishers for content used in its AI answers, on terms publishers have described as opaque.

Pay Per Use is the only one of the three that ties payment to the moment the content does work for the buyer. Whether that is better for a publisher depends entirely on how many fetches a buyer makes per use, which no one outside the buyers currently knows.

What to watch

Which AI companies sign up, and whether any of them publishes its own definition of a use. Cloudflare named no buyers at launch, and the economics of the programme are entirely determined by definitions the buyers write.