AI answer engines consume a publisher's page and return a summary in place of the visit that would have generated revenue. Licensing deals between individual publishers and AI companies won't fill the gap: most sites will never sign one, and no buyer can negotiate separately with millions of domains. What the web lacks is a way to answer an agent with "yes, if you pay." Pay Per Use, now in beta, is one such mechanism.
The design was outlined in July, and since then Cloudflare has worked with buyers and content owners to implement it. A buyer makes an offer for a specific use of a publisher's content, the publisher accepts or declines, the buyer reports every use, and Cloudflare handles billing on one side and payment on the other. Usage and earnings appear in the Cloudflare dashboard; buyers report through a single API.
Why charge for use rather than access
AI systems fetch far more than they ultimately consume — a search engine indexes pages it never surfaces. Billing every crawl forces buyers to pay before they know what their product needs, and many refuse. Tying price to downstream use links cost to the value the buyer receives, which is what makes both sides likelier to show up. Pay Per Crawl, launched in 2025, charges for access; Pay Per Use charges for what happens after access. Publishers pick whichever model fits.
For buyers the appeal is symmetrical. Some of the material their products depend on sits behind blocks or paywalls, and it tends to be the fastest-moving material: news, research, trade publications. With Pay Per Use a buyer can make an offer, pay only for content the product actually uses, be recognized as a verified buyer by every participating publisher, and reach all consenting sites through one API instead of thousands of integrations and one-off contracts. The AI company defines the use it will pay for and sets the price; publishers decide which offers to accept and then see usage frequency and earnings. Enrollment, usage records, billing, and payment are Cloudflare's responsibility.
Mechanics and operator controls
Publishers enroll content for identified AI companies on terms that convert downstream use into revenue, choosing which programs to join while keeping control of crawler access and downstream uses. AI companies identify their crawling activity through Verified bots; where the publisher's controls permit it, they may access and index enrolled content, which can then power a cited answer in AI search, a passage quoted in a research agent's report, a review weighed by a shopping agent, or a recipe adapted by a cooking assistant.
Defining the paid use
An AI company configures its program with Cloudflare by identifying its crawler, describing the use it will pay for, setting a price, and connecting a payment account. A search service might pay when it returns an excerpt from an enrolled page to a customer; a shopping agent might pay when an enrolled review shapes a recommendation — so payment follows use even if the shopper never opens the review. One article can hold different value in different products, and a publisher can accept different offers per use. The buyer proposes the definition; Cloudflare supplies reporting and payment infrastructure.
Accepting or declining offers
Offers surface in the Cloudflare dashboard under Monetize → Pay Per Use, showing the AI company, the use it pays for, and the offered price. Acceptance is the publisher's call, and participation can be ended if an arrangement stops working. No origin change or per-company integration is needed. Program terms also specify what the AI company may do with the content, including any training restrictions.
Reporting a use
A buyer fetches the list of domains that have accepted its offer and then reports each use as one line of JSON containing when it occurred, the URL the content came from, and an event ID.
curl -X POST "https://api.cloudflare.com/client/v4/accounts/$ACCOUNT_ID/pay-per-use/usage-reports" \
-H "Authorization: Bearer $API_TOKEN" \
-H "Content-Type: application/jsonl" \
--data-binary @- <<'EOF'
{"used_at":"2026-09-15T13:42:04Z","url":"https://example.com/article/1","id":"request-001"}
EOFReporting is self-reported: program terms require completeness, and Cloudflare verifies that every reported use maps to an enrolled publisher.
Settlement
Cloudflare aggregates the reported uses, charges the buyer, and pays publishers monthly via the payment account they connected — one integration for buyers, one place to view offers and earnings for publishers.
Usage data and the decisions it informs
Publishers currently see, per AI company, how often their content is used and what it earned, broken out by domain and over time. Business Insights already reports which crawlers visit and what they take; Pay Per Use reports whether the content was in fact used, how often, and for how much. Cloudflare is now working with AI companies to add context around each use — keywords that led to a citation, the topic of the request, or the product it powered — without exchanging personal data. Alongside the Answer Engine Optimization (AEO) tool, which shows how assistants answer questions about a publisher's work, this connects discovery, use, and payment.
Two categories of decisions follow. Commercial: which content earns, which uses are worth it, and whether to stay in a program. Editorial: what to cover, what to refresh, and what to make easier for agents to find.
Beta scope and the agentic web’s economic layer
The beta runs with each buyer and opt-in publisher directly, testing one question: do both sides want to continue? Buyers need content that improves their products at a workable price; publishers need a worthwhile return, trustworthy reporting, and payments that land on time. Longer term the goal is for additional buyers to onboard with their own products and payment models without rebuilding enrollment, reporting, and settlement per publisher. Publishers should eventually get one place to accept or decline offers, counter on price, and price the same content differently by use — last week's reporting may be worth more to an AI product than a decade-old archive page. The beta will determine how to make those choices simple at scale.
Content should reach new audiences and keep earning even when it becomes part of someone else's product: a cited search result, a shopping recommendation, an agent's report. Pay Per Use converts those uses into a commercial relationship — an offer, an acceptance, and a payment with a record attached. Not all content is sold the same way: APIs, tools, and data differ because every request is itself the use, and for those Monetization Gateway, also in beta, charges agents per request over the open x402 protocol. Both share the same foundations of identity, metering, pricing, and analytics. Rather than choosing between blocking every agent and giving work away, publishers get a way to say yes on clear terms, with an account of what happened.
