Cloudflare's Pay Per Use: How Websites Can Finally Charge AI Companies for Their Content

For three years, website owners have watched AI companies take their content and give almost nothing back. Chatbots answer questions using text scraped from blogs and documentation, and the person asking never visits the site the answer came from. The old deal that built the open web, content in exchange for traffic, has been quietly falling apart.

On 30 September 2026 Cloudflare made its biggest move yet to change that. The company launched Pay Per Use, a billing system that pays publishers when AI products actually use their content, alongside a wider rollout of its Monetization Gateway, which lets any site put a machine-readable price tag on any page, API or dataset. Together with a new Auto Router for AI traffic, Cloudflare Pay Per Use amounts to a serious attempt to build a payment layer into the web itself.

Cloudflare sits in front of roughly a fifth of all websites, so when it changes the rules of web traffic, the rules actually change. Let's look at what was announced, how the pieces fit together, and what it means for anyone who runs a website or builds digital products.

From Blocking AI Crawlers to Billing Them

This launch did not come out of nowhere. In July 2025 Cloudflare flipped a big switch: new websites joining its network started blocking known AI crawlers by default, and site owners got a simple dashboard toggle to allow or refuse them. At the same time the company introduced Pay Per Crawl, a marketplace where publishers could charge AI companies a fee every time a bot fetched a page, built on the long-dormant HTTP 402 Payment Required status code.

Pay Per Crawl was a start, but it had an obvious limit. It charged for access, not for value. A crawler might fetch ten thousand pages and use three of them, or fetch one page and quote it in millions of AI answers. Charging the same flat price per fetch fits neither case well. Publishers complained the prices were low, and AI companies complained they were paying for content they never used.

The September launch is Cloudflare's answer to both complaints. Instead of one blunt instrument, there are now several, and they charge at different points in the chain.

What Pay Per Use Actually Does

Pay Per Use, now in beta, moves the billing moment from the crawl to the actual use. An AI company does not pay when its bot downloads your article. It pays when your article shows up in its product: cited in a search answer, quoted in a research report, summarised in a chat response.

The mechanics work in three steps. First, the AI buyer defines what counts as a paid use and what they will pay for it. Second, publishers see those offers in their Cloudflare dashboard and accept the ones they like. Third, the buyer reports usage through an API, sending records with timestamps, source URLs and event IDs. Cloudflare aggregates the reports, bills the AI company and sends publishers a monthly payout.

For a business blog or a documentation site, that means revenue is tied to how useful your content actually is to AI products, not to how often it gets scraped. Well-researched, factual, citable pages become small revenue assets. Thin pages that nobody cites earn nothing, same as before.

The Monetization Gateway: a Price Tag on Any Request

The second piece, the Monetization Gateway, is broader than content licensing. It lets a site charge for any resource behind Cloudflare: a web page, an API endpoint, a dataset, even a tool exposed to AI agents. It runs on x402, an open protocol built around that same HTTP 402 status code.

The flow is surprisingly simple. A client, usually an AI agent, requests a resource. The server replies with 402 and the price. The agent pays, in stablecoins settled directly to the seller's wallet, and repeats the request with proof of payment attached. Cloudflare verifies the payment at the edge before the request ever reaches your server. No checkout pages, no accounts, no invoices, and settlement is designed to complete in under a second.

Cloudflare's own examples give a feel for the pricing: a few cents per search query, a tenth of a cent plus a per-megabyte fee for uploads, 99 cents for a resolved support escalation, a couple of dollars for a generated image. These are micro-transactions that were never practical with card fees, and they only make sense now because the paying customer is software, not a person.

Why This Is Happening Now

The timing is not accidental. Google's AI Overviews and chat assistants have cut referral traffic to many content sites, in some sectors sharply. At the same time agentic traffic is growing fast: AI agents now browse, compare, shop and book on behalf of users, and this autumn's standoff between Amazon and Meta over shopping agents showed how unsettled the rules still are.

Publishers have been pushing back with lawsuits and licensing deals, but both are slow and mostly available to big players only. A major newspaper can negotiate with an AI lab directly; a niche B2B blog cannot. Cloudflare's pitch is essentially collective bargaining at infrastructure scale: the network already sits between AI companies and a large share of the web, so it can set up the toll booths everyone else couldn't.

The Weak Spot: Self-Reported Usage

There is one obvious question about Pay Per Use: the buyers report their own usage. An AI company tells Cloudflare how often it used your content, and your payout depends on that report being honest. Cloudflare sits in the middle, aggregating the reports and billing the buyers, and it can at least cross-check what gets reported against the crawl traffic it already sees. Still, the incentive problem is real, and publishers should watch their early payouts with a somewhat sceptical eye.

The counterargument is that AI companies now have real reasons to take part honestly. Signed usage deals give their retrieval pipelines a legal certainty they currently lack, and a functioning market is cheaper than a decade of copyright litigation. Even so, this part of the system runs on trust plus auditing rather than hard guarantees, and it is fair to treat the first months as an experiment.

What Businesses Building Digital Products Should Do

If you run a content site, start with the basics: check who controls your AI crawler settings. If your site is behind Cloudflare, the controls are in the dashboard today, and joining the Pay Per Use beta is a dashboard form, not an engineering project. Before switching anything on, decide what your content strategy actually is. Blocking every AI bot protects nothing if your competitors' content fills the answers instead; being cited and paid may be worth more than being invisible.

If you publish valuable reference content, invest in being citable. Clear factual claims, named authors, original data and clean page structure make your pages the ones AI products quote. In a pay-per-use world that work converts directly into revenue, not just into rankings.

If you build APIs or SaaS products, the Monetization Gateway deserves a closer look. Usage-priced APIs for agents, per-call tools, paid datasets: things that needed a whole billing team last year can now be an edge rule. We expect agent-facing pricing to become a normal design question in product development, much the way mobile-first became one fifteen years ago.

And if you are adding AI features to your own products, budget for content costs. The free-scraping era is visibly closing, and products that depend on third-party content should model what per-use licensing does to their unit economics before the bill arrives.

Final notes

None of this guarantees the new market works. Prices may settle too low to matter for small publishers, buyers may under-report, and the stablecoin rails will worry some finance teams. But the direction is hard to miss: the web is getting a native payment layer for machines, and content is turning into metered infrastructure.

Our advice is to treat this quarter as the time to get informed and get positioned. Review your crawler settings, look at the beta, and think about which of your pages and APIs are worth a price tag. The sites that do well out of this shift will be the ones that treated it as a strategy question before it became an invoice.