What is agent readiness, and how do you make your website ready for AI agents?
Agent readiness is how well a website lets an AI agent act for a buyer: find a product, read its price and stock, and reach the purchase step without a person stepping in. Axidort measures it as the Axidort Score, from 0 to 100.
Why agent readiness matters
AI assistants and purchasing agents increasingly research and shortlist suppliers before a person visits a website. Gartner expects AI agents to intermediate 90% of B2B buying by 2028, moving more than $15 trillion of spend (Digital Commerce 360). McKinsey estimates agentic commerce could orchestrate up to $1 trillion of US consumer retail revenue by 2030 (McKinsey). Forecasts differ, but they agree on the direction.
An agent can only recommend or buy from a site it can read and use. In the Axidort Index of 50 leading commerce sites, the median Axidort Score is 11 out of 100, 72% of sites score below 50, and 64% challenge or reject an agent before it can finish a purchase (scanned October 1, 2026).
The seven things an AI agent checks
The Axidort Score combines these seven measures. What the agent does carries 90% of the score and file-based signals carry 10%. Steps that do not apply to your kind of site, such as stock on a digital product, are left out and their weight moves to the rest. See the full methodology.
An agent-readiness checklist
- Let declared agents in. Do not serve a CAPTCHA or bot challenge to well-behaved, self-identifying agents. Allow search and answer-engine crawlers in robots.txt, and keep strict rules for abusive traffic.
- Publish price, currency and unit as text. Show the price where a page reader can see it, with the currency and the unit (each, box, case). Avoid prices that only appear after a script runs.
- Add structured data. Mark up products with schema.org Product and Offer: name, SKU or GTIN, price, priceCurrency and availability. Keep it identical to what the page shows.
- Show stock in words. Write "In stock", "Out of stock" or a lead time as text, and mirror it in the Offer availability field.
- Make product pages readable without JavaScript. Render the core product facts in the HTML that the server sends, so a reader that does not run scripts still sees them.
- Keep search and navigation usable without a login. An agent that cannot find a product without an account cannot shortlist you.
- Give checkout a path an agent can reach. Offer guest checkout where you can. If an account is required, make that step clear and standard. Do not let add-to-cart fail silently.
- Provide a sitemap and a product feed. A current sitemap.xml and a product feed give agents a list of what you sell.
- Add llms.txt and machine-readable entry points. llms.txt is a low-cost summary for language models. Support varies, so treat it as an extra. An API or an MCP server is stronger where you can offer one.
- For B2B, expose the business rules. Contract pricing for the signed-in buyer, units and minimum quantities, approved catalogs, purchase-order and quote routes, stock by warehouse, and tax or exemption handling.
How to measure it
Scan your site. An Axidort browser agent searches your site, opens a product, reads price and stock and tries to reach checkout, while a second agent reads the same page the way an AI assistant does. You get an Axidort Score, seven component scores, a grade from A to F and your rank in the Axidort Index. The agent never places an order. Scores are generated by AI agents, so they can change between scans.
Frequently asked questions
01What is agent readiness?
Agent readiness is how well a website lets an AI agent act for a buyer: find a product, read its price and stock, and reach the purchase step without a person stepping in. It is measured here as the Axidort Score from 0 to 100.
02How do I check if my website is ready for AI agents?
Run a scan. The Axidort scanner sends a browser agent through a real purchase journey (search, product page, cart, checkout) and a second agent that reads the same page the way an AI assistant does, then scores seven measures and gives a grade from A to F. It never places an order.
03Does blocking bots in robots.txt or with bot protection hurt agent readiness?
Yes, it can. Agents that identify themselves can be turned away by bot challenges, CAPTCHAs or blanket blocks. In the Axidort Index a bot-protection challenge is the most common reason a well-known site scores low. Allow declared, well-behaved agents, or join a verified-agent programme, and keep rules for abusive traffic.
04Is llms.txt required for AI agents?
No. llms.txt is a plain-text file that summarises a site for language models. Support among AI providers varies, so treat it as a cheap extra signal, not a requirement. Clear prices, stock, structured data and a checkout an agent can reach matter more.
05Do B2B sites need anything different?
Yes. Business buyers need contract pricing, units and minimum quantities, account catalogs, approvals, purchase orders and credit terms, PunchOut or EDI, stock by warehouse and tax handling to be visible and usable by an agent, not only by a person.