AI Readiness Checker: How to Test If Your Website Is Ready for AI Agents

People are starting to hand tasks to AI: comparing products, booking appointments, filling in forms. The assistant doing the work has to understand your website first. An AI readiness checker tells you whether it can. This guide explains what a good AI readiness assessment covers, how to run one on your own site, and how to prioritise the fixes.

What "AI ready" actually means

Being ready for AI has two layers, and most checklists only cover the first one:

• Findable and understandable. AI search systems such as ChatGPT search, Perplexity, Gemini and Copilot can crawl your pages, parse them, and cite them accurately.
• Actionable. A browser agent acting for a user can complete a task on your site, such as searching, adding to a cart or submitting a form, without guessing at buttons and screenshots.

The second layer is new. It is what WebMCP exists for, and it is where almost every site currently scores lowest.

The six-point AI readiness assessment

1. Crawl access (robots.txt)

If your robots.txt blocks AI crawlers, nothing else matters. Check that you are not accidentally disallowing user agents such as OAI-SearchBot, ChatGPT-User, ClaudeBot or PerplexityBot. Training-only crawlers and search or user-triggered agents are separate decisions, covered in our guide to robots.txt for AI crawlers. Also make sure /_next/, /assets/ or other script and style paths are not blocked, or rendering-based crawlers will see a blank page.

2. Structured data

JSON-LD gives machines an unambiguous description of what a page is. At minimum you want Organization and WebSite on the homepage, Article on content, Product and Offer on commerce pages, and FAQPage only where the questions are visible on the page. Accuracy matters more than volume: an agent will trust stale data. See schema markup for AI agents for the specifics.

3. Semantic HTML and accessibility

Agents that read the page rely on the same signals as screen readers: landmarks (main, nav), a sensible heading hierarchy, labelled form fields, real button and a elements, and visible text for important content. If it is a div with a click handler, an agent may not know it is clickable.

4. WebMCP tools

WebMCP lets a page register functions as structured tools an in-browser agent can call. Look for navigator.modelContext registrations or form elements annotated with the declarative attributes. Tools need complete input schemas and descriptions a language model can parse. "Search flights by origin, destination and date" beats "submit". Read the declarative vs. imperative guide to decide which approach suits each page.

5. A machine-readable summary (llms.txt)

An llms.txt file is a short Markdown index of your most important pages. It is cheap to add and harmless, though you should not expect it to move rankings on its own.

6. Performance and stability

Agents time out too. Slow pages, layout shifts that move buttons, aggressive interstitials and bot walls on key flows all cause failures that no markup can fix.

How to run the check in under a minute

1. Open the AgentReady scanner and paste your homepage or a key landing page URL.
2. Wait for the scan. It renders the page in a real browser, so JavaScript-injected tools and schema are detected.
3. Read your Agent Readiness Score and the per-pillar breakdown.
4. Work through the findings in severity order. Fix critical items first (blocked crawlers, missing schema on key pages), then improve tool descriptions and semantics.
5. Re-scan after you deploy, and compare the two reports to confirm the change landed.

How to prioritise what you find

Use this order of operations. It front-loads the changes that remove hard blockers and delays the ones that are mostly upside:

1. Unblock access. robots.txt, bot walls on public pages, broken canonical or noindex tags.
2. Fix accuracy. Structured data that contradicts the visible page is worse than none.
3. Improve semantics. Labels, landmarks and headings are low effort and help humans, search and agents.
4. Add tools. Start with your single most valuable action, usually search or a lead form, and annotate it.
5. Add the extras. llms.txt, sitemaps and monitoring in CI.

Common reasons sites score low

• All the content is rendered client-side with no useful server HTML.
• Forms rely on placeholder text instead of real labels.
• JSON-LD copied from a template that no longer matches the page.
• Tools registered with vague names and no parameter descriptions.
• A CDN or WAF rule that blocks unfamiliar user agents by default.

Make it a habit, not a one-off

Readiness decays. Redesigns, plugin updates and new pages quietly undo earlier work. Run the check after each significant release, and consider adding it to your pipeline so a score drop fails the build. When you hit a good score, you can submit your site to the AI-ready directory so people and agents can discover it.