AI agent readiness checklist: 20 checks before agents come shopping.

This AI agent readiness checklist is the one we use to audit whether AI agents can find, understand and use a website. It has 20 checks in five groups. Work through them in order: there is no point polishing structured data if agents cannot reach the page.

Why an AI agent readiness audit now

Agents are already acting for buyers: Chrome has auto browse, Claude works in Chrome, and assistants fetch pages to answer questions every day. Yet readiness is low. When Cloudflare scanned 200,000 top domains, 78% had a robots.txt, only 4% declared AI usage preferences in it, and 3.9% served Markdown when an agent asked for it.

1. Access

1. robots.txt allows the user-triggered fetchers you want (ChatGPT-User, Claude-User, Perplexity-User) and states your choice on training crawlers.
2. Bot protection and WAF rules let verified and signed agents reach pricing, booking and checkout.
3. No CAPTCHA on the first page view or on simple forms. Save challenges for abuse patterns.
4. Key pages are server-rendered, so content exists before JavaScript runs.

2. Readable pages

1. One clear H1 and descriptive headings on every money page.
2. Prices, plan limits and availability as text, not images, tooltips or calculators. Test with our agent pricing checker.
3. No transparent overlays or hover-only menus covering controls.
4. Chrome Lighthouse's Agentic Browsing category passes on your top pages.

3. Facts

1. Organization, Product or Service, and Offer structured data match what the page says.
2. Price, policies and contact details agree across your site, profiles and marketplaces. Our entity consistency checker compares your pages.
3. Refund, shipping and cancellation policies are on their own crawlable pages.
4. Dates are visible on content that changes, so agents can tell fresh from stale.

4. Actions

1. Every form field has a <label> and a sensible autocomplete value. Check with the agent form checker.
2. Buttons are real <button> elements with clear names, not styled divs.
3. Guest checkout or booking exists, and errors say what to fix in text.
4. Sensitive actions (payment, deletion) ask for a clear confirmation step a person can see.

5. Measurement

1. Server or CDN logs keep the user agent and signature headers. See how to measure AI agent traffic.
2. AI referral traffic has its own channel in GA4.
3. Your three money tasks are tested with a real agent every month, with pass or fail recorded.
4. Someone owns the fix list, and agent results are reviewed alongside SEO.

How to score the checklist

A checklist is a starting point, not a test result. Running a real agent through your real tasks is the only way to know. That is the first step of our AAO service.

Questions people ask

What should an AI agent readiness checklist include?

Five groups: access (robots.txt, bot rules, rendering), readable pages (text prices, no overlays), facts (structured data and consistent details), actions (labelled forms, real buttons, guest checkout) and measurement (logs, GA4 and task tests).

How long does an AI agent readiness audit take?

The checklist itself takes a few hours for a small site. Testing three tasks with real agents and writing the fix list usually takes a few days.

Is there an automated AI agent readiness checker?

Partly. Chrome Lighthouse has an Agentic Browsing category, Cloudflare offers an Agent Readiness score, and our free tools check forms, prices, crawler access and logs. None replaces testing a real task end to end.

Does agent readiness help normal visitors?

Yes. Labelled forms, text prices and real buttons are also accessibility basics, so the same fixes help people using screen readers and keyboards.