AI agents, tools that browse websites and take actions for people, are moving from demos into everyday use. ChatGPT, Gemini, Copilot, Perplexity and browser-based assistants can now read pages, compare options, fill in forms and start bookings or purchases, usually with the person confirming.
When an agent visits your site, it needs to find what it's looking for, understand it and act on it. If it can't, it moves to a competitor's site that's easier to use. Here's a checklist to make sure that doesn't happen.
Access: can agents reach your content?
1. Allow the right crawlers. Check robots.txt doesn't block search engines, AI search crawlers or user-triggered fetchers you want. Use our AI crawler check and robots.txt tester.
2. Don't over-block at the firewall. Bot protection that challenges every automated visitor can block legitimate agents too. Review your CDN's bot settings. See bot traffic now outnumbers humans.
3. Make key pages indexable. No accidental noindex or wrong canonicals. Check with our indexability checker.
4. Keep content in the HTML. Prices, product details, opening hours and policies should be in the server-rendered HTML, not only loaded by JavaScript after interactions. Read JavaScript SEO.
5. Publish an llms.txt file. A short guide to your most important pages for AI tools. See llms.txt explained, or create one with our free llms.txt generator.
Structure: can agents understand your pages?
6. Use proper headings in order. One H1, then H2s and H3s that describe each section. Check with our heading outline tool.
7. Use semantic HTML. Real buttons for actions, real links for navigation, lists for lists, tables for tabular data.
8. Add structured data. Organization, LocalBusiness, Product, Offer, Service, Event, FAQ and Article markup tell agents exactly what things are. Use our schema generator and validate with the JSON-LD checker.
9. Keep a clean accessibility tree. Agents rely heavily on the same information screen readers use: labels, roles and names. Follow our WCAG 2.2 checklist.
10. Avoid layout shifts and pop-up walls. Interstitials, cookie walls and elements that move around confuse agents as much as people.
Content: can agents find the facts they need?
11. State key facts plainly. Prices or price ranges, sizes, compatibility, delivery times, service areas, opening hours.
12. Make policies easy to find. Delivery, returns, cancellation, warranties and privacy, linked from every relevant page.
13. Answer common questions on the page. Use FAQ sections based on real customer questions. See FAQ pages that get cited by AI.
14. Keep information current. Stale prices and old opening hours lead agents to give wrong answers about you.
Actions: can agents complete tasks?
15. Label every form field. Visible labels connected to inputs, clear required fields, sensible input types. Read form design best practices.
16. Write helpful error messages. "Enter a UK postcode, like SW1A 1AA" helps agents and people correct mistakes.
17. Avoid unnecessary steps. Guest checkout, short booking flows and no forced account creation.
18. Use standard payment methods. Express wallets and widely supported payment providers work better with agent flows than unusual custom checkouts.
19. Consider WebMCP. The proposed standard lets pages declare actions like "check availability" directly to agents. It's still in testing, but worth prototyping on one simple form. Read WebMCP explained.
Trust: will agents (and people) trust what they find?
20. Show who you are. Company name, address, contact details, reviews and clear ownership. Agents weigh trust signals when recommending businesses, just like people. See E-E-A-T explained.
A quick self-test
Ask an AI assistant to do a real task on your site: "Find the price of [service] on [your site] and book the earliest appointment" or "Check whether [product] comes in size 10 and how long delivery takes." Watch where it struggles. Every hesitation is a fix.
Example: making a booking flow agent-ready
A physiotherapy clinic's booking page before and after:
| Before | After |
|---|---|
| Prices only in a PDF | Prices for each treatment in the HTML, with Service structured data |
| Availability loaded in a third-party iframe with unlabelled buttons | Booking widget with labelled date, time and treatment fields, or a booking page agents can read |
| "Call us to book" for new patients | Online booking for new and returning patients, with clear first-visit information |
| Cancellation policy hidden in terms | Cancellation policy summarised next to the Book button |
| Opening hours only in an image | Opening hours in text and LocalBusiness structured data |
| CAPTCHA on every step | Bot protection that doesn't block genuine users (and agents acting for them) unnecessarily |
Now when someone asks an assistant to "book the earliest sports massage at a clinic near me this week under £60", the clinic's details are readable, comparable and actionable. Without those changes, the assistant moves on to a clinic whose site it can use.
Scoring yourself
| Score | Meaning |
|---|---|
| 18–20 | Agent-ready; keep monitoring |
| 13–17 | Mostly ready; fix the gaps in forms and structured data |
| 8–12 | Agents will struggle with key tasks |
| Under 8 | Start with access, headings, labels and key facts in HTML |
The short version
An agent-ready website lets AI agents reach, understand and act on your content: allow the right crawlers, keep key facts in HTML, use semantic structure and structured data, label forms clearly, simplify flows and show trust signals. Most of the work also improves SEO and accessibility.
Want a full agent-readiness audit? Run a free AI Website X-Ray, or ask our web development team to review your site.




