On Cloudflare’s Q3 earnings call, CEO Matthew Prince called the company’s help for customers who want more agent traffic “sort of the new version of SEO.” He said Cloudflare tells OpenAI when a webpage has changed, such as a restaurant updating its prices, so OpenAI’s crawler doesn’t re-fetch pages that are the same. He said OpenAI pays for this and he expects other AI companies to follow. He also expects agents that explain why they are visiting to pay less for access. Search Engine Journal reported it on 8 November 2026.

What the claim gets right

Agents do visit sites differently from people. Prince’s example was an agent pulling menus from 1,000 restaurant sites for one diner. One restaurant gets the booking and 999 carry the cost. A human might look at two or three pages. An agent works at a much larger scale.

So a feed that says “this page changed” saves everyone work. The crawler skips unchanged pages and the site serves fewer pointless requests. Cloudflare also lets customers choose between welcoming agents and blocking them, and customers who block them aren’t charged for that traffic. I see a real need there.

What the article doesn’t say

The article is clear about its gaps. Prince gave no Act 4 revenue figure. He didn’t say what OpenAI pays or when payments began. The call didn’t address whether sites in the pilot get a share. He also didn’t mention a Cloudflare service that optimises a website for agent traffic.

The micropayment idea is a prediction. He said he doesn’t know how long it will take to become a major revenue source. The products behind it are in beta, and Pay Per Crawl is listed as a private beta.

So “new SEO” describes where Cloudflare wants to sell. It tells you nothing yet about what makes an agent recommend your business.

Switch on the cheap layer and move on

Change hints, crawler controls and clear access rules cost little and do no harm. If your platform or host offers them, turn them on. If they can be automated, automate them and stop thinking about them.

I treat llms.txt and the “AI-ready” checklists the same way. AI search engines are smarter than to put much weight on a simple file. But if something is cheap, harmless and automatable, you just do it. It isn’t the strategy.

This also helps with a worry I hear from small operators. Many feel they must pick a side now, either open the doors to every agent or block them all. The big platforms can afford to wait and see how agentic commerce turns out, and some may block agents to keep their hold on shoppers. A small business can’t wait. Its shift is already decided, and the decision is to get data-centric.

What decides whether an agent names you

A freshness signal tells a crawler that your page changed. It says nothing about whether the page is worth quoting.

When someone asks an agent for the best option, the agent picks the business it understands most clearly and can quote most precisely. Winning that recommendation takes enough attributes that are materially and measurably different from your competitors’. Numbers survive. Adjectives get dropped.

Take a clinic that updates its prices. The change hint tells OpenAI the page moved. If the page says “affordable, trusted care”, the agent has nothing to quote. If it lists the price of each treatment, the opening hours, which insurers are accepted and how soon the next appointment usually is, the agent has facts to repeat. The same goes for a shop. Sizes, materials, delivery times and return terms get quoted. “High quality” doesn’t.

Your website is also only one surface of many. Wherever your data shows up, it needs the same precision and an accurate reading of what you are.

People also ask their agents the same question in slightly different ways. Your information has to cover a much broader spread of questions, with varied and detailed attributes. AeroChat does this by publishing many pages, each aimed at one integration with specific software, so each page catches one long-tail question.

What a business should do

  1. Turn on whatever change hints and crawler controls your host or CDN offers. Decide whether you welcome agents or block them, set it once and leave it.
  2. Go through your key pages and replace adjectives with facts: prices, specifications, times, terms, locations.
  3. Find what is materially different about you compared with your competitors and state it with numbers.
  4. Cover the long tail. Write a page for each specific question a customer might put to an agent.
  5. Check your pages on a schedule and keep them current. A fresh signal on a stale page only tells the agent faster that your facts are wrong.
  6. Test by asking the agents. Record which ones name you and on what date, because the answers change week to week.

Source: Search Engine Journal: Cloudflare CEO Likens AI Agent Optimization To A New Version Of SEO