Build1 publisher2 min readPublished
Cloudflare's Pay Per Use beta bills AI buyers on usage they report themselves
Cloudflare's Pay Per Use beta charges AI companies for each use of a publisher's page they report, one three-field JSON line per event. Publishers can license without negotiating, though the buyer defines the use, sets the price and keeps the count.
The Engineer · Build desk
Drafted by a language model from the sources cited here and checked against its claim ledger before publication. How we use AISend a correction
What happened
- Cloudflare first outlined Pay Per Use in July and says it has worked with buyers and content owners since then to build it.
- Offers appear in the Cloudflare dashboard under Monetize > Pay Per Use, listing the AI company, the paid use and the price, and publishers can accept or later stop participating.
- Each program's terms set what the AI company may do with the enrolled content, including any restrictions on training.
- Pay Per Crawl, launched in 2025, still charges for access, and Cloudflare says publishers can pick whichever of the two models suits them.
Compiled by The EngineerSomething wrong?How this is made
Why it matters
- exposure A publisher's income depends on an event log kept by the party paying the bill, recorded inside a product the publisher cannot observe.
- constraint A small publisher's only leverage over price is refusing or leaving an offer, since the buyer writes both the use definition and the price.
- capability An AI company can reach blocked or paywalled news, research and trade content through one API as a verified buyer, without per-site contracts.
- decision Choosing between Pay Per Crawl and Pay Per Use means deciding whether fetches that never become uses should earn a site anything.
A billable event starts in the buyer's code [4]. The buyer pulls the list of domains that have accepted its offer. It then posts each use as one line of JSON with a timestamp, the source URL and an event ID [4]. Cloudflare's sample request sends those lines as application/jsonl to an account-scoped usage-reports endpoint, authenticated with a bearer token [5]. Cloudflare then bills the buyer and pays the publisher [3].
Cloudflare's post says it plainly: "Usage is self-reported" [6]. The fetch is identified, because AI companies declare their crawlers through Verified bots [15]. The use comes later, inside the buyer's product. A search service could pay when it returns an excerpt from an enrolled page [10]. A shopping agent could pay when an enrolled review shapes a recommendation, even if the shopper never reads the review [9]. The only record of either event that the post describes is the buyer's report [2]. In billing terms, the customer reads its own meter [2]. The available text does not show whether Cloudflare checks those reports against its crawl records [6].
The pitch to publishers is about scale. Cloudflare argues that most publishers will never sign a licensing deal with the companies using their work, and that no company can negotiate with millions of sites [17]. Pay Per Use solves that by letting one side write the terms [1]. Each AI company defines the use it will pay for and sets the price [7]. Cloudflare wrote that "The buyer proposes the definition of the use being paid for" [8]. A publisher can accept, decline, or walk away later [1].
The integration work is well judged on both ends. A publisher makes no origin change and builds no technical integration with each AI company [21]. The report format is close to minimal: three fields a line, sent as JSON Lines in one POST [4][5]. The event ID on every line gives each billed use its own key [4].
The case for moving the charge, according to Cloudflare, is that AI products fetch far more than they use [14]. It says charging per crawl makes buyers pay before they know what they need, and many won't [14]. It hopes paying for use brings more buyers to the table and more money to publishers [18]. For a site to come out ahead, the per-use price times the uses buyers report has to beat the crawl fees it would have collected from buyers willing to pay them [3].
What to watch
- Whether Cloudflare documents how it reconciles buyer usage reports against Verified bot crawl records, or gives publishers audit rights.
- Which AI companies enroll as buyers during the beta, and what per-use prices show up in dashboard offers.
- Whether publishers gain a way to counter-propose price or use definitions, or the flow stays accept-or-decline.