Your website needs a machine price sheet

Yesterday the question about your website was fairly simple. Which AI systems are allowed through the door?

Search crawlers, training crawlers, answer engines, buyer agents, browser agents. Different visitors, different motives, different levels of risk. The old “allow bots” or “block bots” switch is too blunt for any of that now.

But there is a second question sitting right behind it, and it is more commercial. When the machines get through the door, what do they have to pay for?

That sounds odd if you still picture a website as a stack of pages for humans. A person lands, reads, fills in a form, maybe buys. Search engines crawl the public bits and send a little traffic back. That was the trade, and everyone understood it.

AI has made the trade messy.

An agent does not browse like a person. It may not look at your ads. It probably does not care about your navigation. It will not visit ten pages and slowly warm up to your brand story. It usually needs one thing: a price, a data point, a quote, a stock check, a product spec, an appointment slot, an API response, a tool call, a document, a calculation. It takes the thing it came for and moves on.

Sometimes that is exactly what you want. If a buyer’s agent checks whether you serve Birmingham and then brings you a qualified lead, lovely. If an answer engine reads your service page and cites you properly, fine. If a customer’s agent books a call through a safe workflow, even better.

But if an AI system hammers your content, reuses your original work somewhere else, drains your API, calls your tools, skips your checkout, ignores your terms and sends you nothing back, that is not marketing. That is extraction with nice branding.

This is why Cloudflare’s latest move is worth a proper look.

The headline piece is the Monetization Gateway. Cloudflare says it will let customers charge for web pages, datasets, APIs and MCP tool calls sitting behind Cloudflare, using x402-style machine payments. In plain English, a machine asks for a resource, the server can answer “payment required”, the machine pays, and the request carries on. No normal checkout. No human typing in card details. No long procurement dance.

It is early. It is on a waitlist. The payment side leans on stablecoins, which will make plenty of ordinary business owners roll their eyes, and not unfairly. There are accounting, legal, tax and fraud questions that no tidy protocol diagram is going to solve this year.

The direction still matters. Cloudflare is no longer just talking about blocking AI bots. It is talking about classifying them, measuring them, charging them, and letting resources on the web behave more like billable infrastructure. That is the useful signal. The agent web will not only be a visibility problem. It will be a pricing problem.

Most companies have not even mapped their assets properly.

They have public pages, PDFs, blog posts, calculators, product feeds, FAQs, appointment systems, quote forms, support knowledge bases, pricing tables, case studies, training material, API endpoints, internal dashboards, and maybe, soon, MCP tools. To a human, those feel like different corners of the website or the business. To an agent, they are all just resources.

Some of those resources should be free. You want machines to understand what you sell, where you operate, what problems you solve, what proof you have, and how a buyer can engage. That is discovery. Hide all of it and you make yourself harder to find.

Some should be public but controlled. A service page can be indexed and referenced, but you may not want it reproduced wholesale or hoovered up for model training. A product feed can help buyer agents compare you, but the data still needs to be accurate, current and attributed.

Some should be paid. If an agent wants a premium dataset, a high-volume API call, a specialist calculation, a proprietary benchmark, a rendered report, or a tool invocation that costs you money every time it runs, there is no law that says it has to be free.

Some need identity before price. If an agent wants to book, cancel, refund, change, submit, sign or buy, the first question is not “what does it cost”. It is “who is this agent acting for, and is it allowed to do that”.

And some resources should not be machine-accessible at all. That is the bit the hype crowd skips because it ruins the mood. Tough. A sane business does not expose every useful internal function to the open web because someone said the word agent in a keynote.

So the next website brief needs a machine price sheet.

Not a literal public menu for every bot on earth. More like an internal map of your own business. This content is free for discovery. This content can be referenced but not reproduced. This data is available to verified partners. This endpoint is paid per request. This tool call requires authentication. This action needs human approval. This material is blocked from training. This system is not exposed to agents at all. That is not glamorous web design language. It is better than that. It is useful.

A normal business owner does not need to wake up tomorrow and wire in x402 payments. Most should not. There is no prize for being first to bolt stablecoin micropayments onto a half-understood website while your basic lead forms are still broken. But you do need to start thinking differently about what your site actually contains.

A website used to be a shop window. Then it became a conversion machine. Then it became a content and SEO asset. Now it is becoming a set of resources that humans, search engines and agents all use in different ways.

If you run an agency, that changes the work. Building pages is not enough. Writing SEO copy is not enough. Sticking a chatbot in the corner is definitely not enough. The new work is mapping the commercial surface of the business. What can be found. What can be trusted. What can be called. What can be bought. What can be automated. What can be charged for. What can be rolled back. What must be logged. What needs a human in the loop. That is a better website conversation than “do we rank in ChatGPT”.

The two lazy reactions are easy to predict. One camp leaves everything open because they are terrified of going invisible. They call it openness until they notice their best material is being consumed with nothing coming back. The other camp blocks everything because they are terrified of being scraped. They call it control until they realise buyers now use agents and answer engines as their first filter. Neither is a strategy.

The practical route is duller and better. Classify the machine interaction. Is it a lead. Is it a cost. Is it a risk. Is it a sale. That one question changes the brief. If it is a lead, make the information clean, structured and easy to cite. If it is a cost, meter it or throttle it. If it is a risk, require identity and approval. If it is a sale, design the transaction properly and keep the receipt.

This is where the post-agency model gets interesting. The next useful marketing company is not the one producing more output. It is the one helping businesses turn their websites and commercial systems into controlled operating layers. Human-readable where humans matter. Machine-readable where discovery matters. Machine-actionable where it is safe. Machine-billable where there is real value. And locked down where the exchange is not fair.

That is the part I would start planning now. Not because every business needs agent payments this quarter, because they do not. Because the web is moving from pages and visitors to resources and requests. And if a machine can use your work, you have to decide whether that use is marketing, service, risk, or revenue.

Right now, most businesses have not decided. That is the opportunity.

---

Jason Sibley is the founder of Cleo, a post-agency marketing and AI company. JasonVsTheNoise is where he writes about what is actually happening with AI, marketing, and how businesses should be thinking about both.