AI agents
AI agent readiness checklist for your website
AI agent readiness is whether an AI agent can reach your site, read what you sell, and complete the next step for the person it works for. Use this checklist to find the gaps, test them with free tools, and fix them in the order that matters.
Short answer
AI agent readiness measures whether AI agents acting for a person can use your website: reach the pages, read your products or services, prices and policies as text, and complete actions such as a form or a checkout. Check it with a seven-area checklist, from access and structured data to measurement, and fix whatever blocks agents completely first.
- Agents read a page from screenshots, the HTML and the accessibility tree, and each is a place a page can lose them.
- The checklist covers seven areas: access, readable facts, structured data, actions, policies and trust, optional guidance files and measurement.
- Put the price, delivery cost and returns terms in the HTML as text, and make structured data match the visible page.
- Fix what blocks agents completely first; optional extras such as llms.txt come last.
- You can test readiness yourself with the page source, Lighthouse, Cloudflare's scanner, a rich results test and an AI assistant.
What is AI agent readiness?
AI agent readiness is how well your website works for AI agents acting on someone's behalf: whether they can reach your pages, read what you sell, the price and the policies, and complete the next step, such as a quote form or a checkout. A site can convert people well and still fail an agent on every one of those points.
The phrase is also used for two other things: whether a company's own systems are ready to run AI agents, and whether a code repository is ready for AI coding agents. This page is about neither. It is about your public website, and the agents your customers send to it.
How AI agents read a website
AI agents read a website in three ways, according to Google's web.dev guide to agent-friendly sites: from screenshots, from the HTML, and from the accessibility tree, the browser's summary of every element's role, name and state. Modern agents combine them. Each is a place where a page can lose an agent.
| Method | What the agent gets | What breaks it |
|---|---|---|
| HTML | The page structure, text and attributes, and which button belongs to which product | Facts that only appear after scripts run, prices in images, generic div elements used as buttons |
| Accessibility tree | Roles, names and states of interactive elements, without the visual noise | Unlabeled fields and buttons, invalid roles, content hidden from the tree but still clickable |
| Screenshots | Layout, grouping and visual emphasis | Transparent overlays, tiny controls, layouts that shift while the page loads |
| Structured data and feeds | Product, offer, shipping and return facts in a fixed format | Markup that disagrees with the visible page, stale feeds |
The AI agent readiness checklist
This checklist groups the checks into seven areas: access, readable facts, structured data, actions, policies and trust, optional guidance files, and measurement. Each line is a yes or no question. Work through it on your most important page types first: a product or service page, the pricing page, and each step of checkout or the lead form.
1. Access: can agents reach your pages?
- robots.txt states a rule for the AI crawlers you care about, rather than relying on defaults you have never checked.
- Your CDN, firewall or bot protection does not block the AI crawlers and agents you have chosen to allow.
- robots.txt points to your sitemap, and the sitemap lists your key pages.
- Key pages load without a login, a country selector or a full-screen popup in the way.
- Pages you do not want used, such as account or cart URLs, are excluded on purpose.
2. Readable facts: can agents read what you sell?
- The product or service name, price and currency are in the HTML your server sends, as text.
- Variants such as size and color, and stock for each, are readable without clicking a picker.
- Delivery cost and delivery time are stated on the product page or one link away.
- Key specifications are in text or a real table, not only in images or a PDF.
- The same price appears on the page, in the structured data and in your product feed.
- Each page has one clear h1 and headings that describe what follows.
3. Structured data: does the markup match the page?
- Product pages carry Product and Offer markup with price, currency and availability.
- Shipping details and a merchant return policy are marked up, or your feed supplies them.
- Organization markup names your business, logo and contact details.
- Every marked-up fact is visible on the page in the same words and numbers.
- There is no rating or review markup you cannot back with real, visible reviews.
- Pages pass Google's Rich Results Test without errors.
4. Actions: can an agent complete the next step?
- Buttons are
buttonelements and links areaelements, not styleddivorspanelements. - Every form field has a label tied to it with the
forattribute. - No transparent overlay sits on top of an interactive element.
- The layout does not shift after load, and key buttons sit in the same place across similar pages.
- Guest checkout or a short quote form exists, without forcing account creation first.
- Error messages explain in text what went wrong and how to fix it.
5. Policies and trust: can an agent confirm the terms?
- Shipping, returns and guarantee terms are short, specific statements on a crawlable page.
- Each product page links to those policies with descriptive link text.
- Contact details and the business's legal name and address are on the site as text.
- Any claim an agent might repeat, such as "free returns", says what it covers and for how long.
6. Optional guidance files
- An llms.txt file summarizes the site and links to its key pages, if you choose to publish one.
- Your ecommerce platform or payment provider has told you which agent checkout protocols, such as ACP or UCP, it supports.
7. Measurement: will you know if it works?
- Your analytics shows referral visits from AI assistants as a separate segment.
- Your server or CDN logs are checked for the published AI user agents.
- Google Search Console's generative AI performance report is part of your monthly review.
- Conversion rate is tracked for AI-referred visits, not just visit counts.
How to test your site's AI agent readiness
You can test AI agent readiness yourself with free tools: read the page source, run Lighthouse's experimental Agentic Browsing category, scan the site with Cloudflare's agent readiness checker, validate structured data, and then ask an AI assistant to find your facts. Repeat it on each important page type.
-
Read the page source
Open a product or service page, view the source, and search it for the price, the delivery cost and the word "return". If a fact is not in the source, some agents and crawlers will not see it.
-
Run Lighthouse's experimental agent audits
Chrome's Lighthouse has an experimental Agentic Browsing category that needs Chrome 150 or later. It reports how many agent checks a page passes, including names and labels in the accessibility tree, layout shift and llms.txt, rather than a score out of 100.
-
Scan the site at isitagentready.com
Cloudflare's free scanner scores discoverability, content, bot access control and capabilities, and also reports whether a site supports the UCP and ACP commerce protocols, which do not count toward its score.
-
Validate your structured data
Run key pages through Google's Rich Results Test and compare every value it finds with what the page shows a visitor.
-
Ask an AI assistant
Give an AI assistant your page URL and ask for the price, delivery time and return policy. Note what it gets wrong or cannot find. Answers vary between runs, so repeat the question and treat it as a spot check, not a measurement.
-
Check your logs
Look for AI user agents in your server or CDN logs and for blocked requests from them in your firewall reports.
AI agent readiness tools compared
Free agent readiness tools each look at a different layer: Lighthouse at the page and its accessibility tree, Cloudflare at site-wide standards and protocols, Google's Rich Results Test at structured data. None reads your page as a buyer would, so use them together.
| Tool | What it checks | Good for |
|---|---|---|
| Lighthouse Agentic Browsing | Accessibility tree names and roles, layout shift, llms.txt, WebMCP tools (experimental) | Whether an agent can use a specific page |
| Cloudflare isitagentready.com | robots.txt, sitemap, Markdown for agents, bot access rules, API and protocol discovery; reports UCP and ACP support | Site-wide standards and how you compare |
| Google Rich Results Test | Structured data on a page and its errors | Product, offer and policy markup |
| Convertica free CRO audit | Eight checks for people and AI agents, including whether an agent can read what the page sells, the price and the policies | Agent readiness alongside conversion problems for people |
Most sites are early. When Cloudflare scanned a filtered set of the 200,000 most visited domains for its April 2026 launch, 78% had a robots.txt file but most were written for search engine crawlers rather than AI agents, 4% declared AI usage preferences with Content Signals, and 3.9% served Markdown on request (Cloudflare, 2026). Being early is an advantage if you fix the basics now.
Which agent readiness problems to fix first
Fix agent readiness problems in order of how completely they block an agent: first anything that stops an agent reaching the page, then missing or conflicting buying facts, then actions an agent cannot complete, and only then optional extras such as llms.txt or Markdown versions of pages.
| Priority | Problem | Why it comes here |
|---|---|---|
| 1 | Agents blocked by robots rules or bot protection | Nothing else matters if the page never loads |
| 2 | Price, stock, shipping or returns missing from the HTML | The agent cannot check you against the buyer's rules |
| 3 | Structured data or feed that disagrees with the page | Conflicting facts cost trust with agents and search engines |
| 4 | Unlabeled forms, fake buttons, overlays, shifting layout | The agent finds you but cannot finish the task |
| 5 | Vague policies | The agent cannot quote terms with confidence |
| 6 | No llms.txt, no Markdown versions, no protocol support | Useful for some agents and platforms, but optional today |
For why this matters to sales, and the protocols behind agent checkout, read the guide to AI agents and commerce. Convertica's own site is built to this checklist: content in the HTML on load, AI crawlers named in robots.txt, one structured data graph per page built from the visible content, and a generated llms.txt.
Get your AI agent readiness checked
AI agent readiness is one of the eight checks in Convertica's free CRO audit, an app you can run now: enter your website and email, and watch it run live on your page. If you want help deciding what to fix and in what order, CRO advisory is led personally by Kurt Philip, with Convertica's CRO team, and is priced after your audit. Your team makes the changes, or, with full implementation, Convertica's team builds the fixes from your audit.
AI agent readiness FAQ
What is AI agent readiness?
AI agent readiness is how well a website works for AI agents acting for a person: whether they can reach the pages, read the product or service facts, prices and policies, and complete actions such as a form or a checkout. It is checked separately from how the site works for people, although most fixes help both.
How do I check if my site is ready for AI agents?
Start by viewing the page source of a key page and confirming the price, stock, shipping and returns are there as text. Then run Chrome Lighthouse's experimental Agentic Browsing category, Cloudflare's free scanner at isitagentready.com and Google's Rich Results Test. Finally, ask an AI assistant to find your product's price and return policy from the URL alone.
Is AI agent readiness the same as SEO?
No, but they overlap. SEO is about being found and ranked in search. Agent readiness is about what happens when an agent reads your page or tries to act on it: can it confirm the facts and finish the task? Crawl access and structured data matter to both.
Do I need an llms.txt file to be agent ready?
No. An llms.txt file is optional. Lighthouse marks its check as not applicable when the file is missing, and Google says its Search features do not use it. It is cheap to add, but readable facts on your pages matter far more.
Does AI agent readiness help human visitors?
Usually, yes. Visible prices, plain-text policies, labeled form fields, real buttons and a stable layout help people on phones and people using assistive technology as much as they help agents. Google's web.dev guidance makes the same point.
Does the free CRO audit check AI agent readiness?
Yes. AI agent readiness, whether an agent can read what a page sells, the price and the policies, is one of the eight checks in Convertica's free CRO audit, an app you can run now. In CRO advisory, Kurt Philip reviews it with you.