Cloudflare Pay Per Use beta — publishers paid when AI *uses* content

Cloudflare opened Pay Per Use in beta (blog 2026-09-30): a content-network path so publishers can say “yes, if you pay” when AI products use their work—without each site negotiating a bespoke licensing deal. Buyers offer a price for a defined use; publishers accept or decline; Cloudflare bills the buyer and pays the publisher (blog).

This is a Desk Bot tools/agents briefing. HARD fence vs T6 Monetization Gateway / x402: Gateway is seller-side HTTP 402 metering for APIs, MCP tools, and data where every request is the use. Pay Per Use is the publisher-payout content network—verified buyers crawl under publisher controls, then self-report downstream uses; Cloudflare settles monthly. Dashboard path Monetize → Pay Per Use ≠ Monetization Gateway. Also distinct from Pay Per Crawl (access charge). Do not rehash Auto Router, cf CLI, or Containers.

Pay for use, not the crawl

Pay Per Crawl (2025) charges for access; Pay Per Use pays for downstream use—e.g. an excerpt in AI search, a review shaping a shopping recommendation, a passage in an agent report. Publishers can choose the model that fits; the same article can earn under different buyer-defined uses (blog, July plan).

Offer → opt-in → report → settle

1. AI buyer sets up a program (identifies crawler, defines paid use, sets price, connects payment).
2. Publishers review offers under Monetize → Pay Per Use (company, use definition, price) and accept/stop at will—no origin change or per-buyer technical integration.
3. Buyer fetches accepted domains and reports each use.
4. Cloudflare aggregates, charges the buyer, pays publishers monthly via connected payment account (blog).

AI companies identify crawling via Verified bots; publishers retain crawler/access controls and choose which programs to join. Program terms can restrict what buyers may do with content, including training (blog).

Soft: self-reported usage (trust model)

Buyers POST usage as JSONL lines (used_at, url, id) to …/pay-per-use/usage-reports. Blog: usage is self-reported; program terms require complete reporting; Cloudflare checks each reported use maps to an enrolled publisher (blog). Completeness/accuracy rests on buyer self-reporting + enrollment checks—not independent third-party metering of every AI answer; do not imply cryptographic proof of each citation.

Soft: beta scope + visibility

During beta Cloudflare works directly with each buyer and opt-in publishers. Do not invent open-beta/GA dates, named settlement SLAs, or claim July pilot partners (e.g. Ceramic.ai / You.com from the plan post) are confirmed live payout customers in the Sep 30 launch post unless re-verified (blog, July plan).

Publishers see reported uses and estimated earnings by buyer/domain over time; blog ties that to Business Insights and Answer Engine Optimization (AEO) as complementary views. Earnings/usage charts and “next” richer per-use context are vendor-described / beta—do not invent GA dashboards or named live buyer payout figures beyond the Sep 30 post (blog).

Who should care

Publishers and AI product teams negotiating content use for answers/agents should start at the Pay Per Use blog—keep beta + the self-report trust model, leave HTTP 402 API metering to Monetization Gateway, and don’t equate this with Pay Per Crawl access pricing.