Go-to-market

Suganthan Mohanadasan set his site to charge AI agents one cent per page, and Claude Code paid it from a wallet during a task

September 17, 2026 at 12:30 PM PT

Hero image for Suganthan Mohanadasan's blog post about charging AI agents one cent per page using the x402 protocol

Image: Suganthan Mohanadasan

Why it mattersA publisher's own server can now hold a page back until an agent pays and can check the payment itself, which is the crawl model teams have to decide about before Cloudflare turns it on in a dashboard.

Suganthan Mohanadasan, co-founder of Snippet Digital, published a working setup on 15 September that charges AI agents one cent per page in testnet USDC. His own site has been running the code since 8 August, and he reports that five payments settled on 15 September on the Coinbase-run x402 facilitator, with a public transaction hash for each. One of them was made by Claude Code, running on his own machine from a wallet he had set up for it, during an ordinary web-fetch step in a task.

What is running on the site

The site sits on a Cloudflare Worker built with the open x402 protocol. When an agent requests a paid page, the server returns HTTP 402 with a payment-required header describing the price and payment address. A compatible client signs a transaction, and the Coinbase facilitator at x402.org verifies and settles it on-chain. Every paid crawl has a public receipt on the site's earnings page. Suganthan states that no ordinary search crawler is paying, and all payments so far have come from his own agents on testnet USDC, which has no monetary value.

Why this exists now

Google is running an AI contribution pilot that pays some publishers when Google decides their content contributed to an answer in AI Overviews, AI Mode or the Gemini app. Suganthan cites Search Engine Roundtable and Digiday, both from 14 September. Publishers get a monthly earnings figure in Search Console with no breakdown of how it was calculated.

The x402 route sits on the other side of that trade. The publisher sets the price before an answer is produced, the agent decides whether to pay, and the transaction is on-chain. Suganthan writes that with a Google-style pay-per-use pilot he is trusting the platform's number, and with pay-per-crawl he can check the request and the payment himself. Cloudflare is building a Monetization Gateway around the same x402 idea, and Suganthan says his demo runs independently of Cloudflare's closed Pay Per Crawl programme.

The x402 protocol has organisational backing beyond Cloudflare and Coinbase. Suganthan quotes the Linux Foundation announcement that the x402 Foundation launched operationally on 14 July with 40 members, including Visa, Mastercard, Google, AWS and Stripe.

What the payment loop looks like in practice

During one live session Claude Code hit the 402 response through its WebFetch, a hook signed a payment worth one cent in testnet USDC, and the agent then quoted the settlement transaction in its own summary of the task. The transaction hash is what makes this different from a metered feed number: the publisher and the agent both see the same on-chain record.

Real search crawlers are not doing this yet. Suganthan is explicit that GPTBot, ClaudeBot and Googlebot will not pay a price today, and a site that switches paid gating on now will lose the citations it currently earns from those crawlers. The point of the setup is that the loop works end to end before it becomes another toggle in a Cloudflare dashboard, and a publisher who has run it once knows what the trade will actually look like.

For a team deciding what to do about AI crawlers, the immediate question is whether to keep the page free and be cited in an answer, or to gate it and be paid for a request that may not arrive. The setup Suganthan documents lets that question be tested against real transaction data rather than a vendor's monthly figure.

Source

Reported by: Search Engine Journal

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.

More from AI News

Cloudflare adds a Disallow AI Training setting that keeps Google, Apple and Bing crawling for search

Cloudflare has split its bot controls so a site can opt out of AI training while keeping Googlebot, Applebot and Bingbot crawling for search, reversing an earlier plan that would have blocked all three from Cloudflare-fronted sites that refused training.

Source: PressGo-to-market

SEJ column shows the AI crawl-to-refer ratio for Anthropic has been quoted at seven different values in 13 months, all from Cloudflare

Duane Forrester traces seven different versions of Anthropic's crawl-to-refer ratio, all attributed to Cloudflare and published inside 13 months, and shows the metric has four hidden denominators that never travel with the number.

Source: PressGo-to-market

Claude and Claude Code search the web at different rates and mention different brands, one study finds

A Profound study of 1,724 prompts run through both Claude and Claude Code found the two products search the web at very different rates and rarely surface the same brands.

Source: PressGo-to-market