What makes a site agent-ready?

Reliable, structured data, clean feeds, machine-readable terms of sale, and an exposed agent endpoint such as a UCP profile, an MCP server or an llms.txt file.

Where should I start if my score is low?

With family A, product data: everything else builds on it. Make the repository reliable before structuring, and structure before exposing.

Agent-ready checklist: is your site ready for agentic commerce? (20 points)

How to use this

Twenty points, five families, following the order of the five workstreams. Score one point per box ticked: 0–7, foundations are your subject; 8–14, you're visible but not yet buyable; 15–20, you're agent-ready — move on to optimisation.

Staircase diagram of the five preparation workstreams, matching the five families of the agent-ready checklist. — Fig. — The checklist's five families follow the workstream order.

A. Product data (the foundation)

1. A single source of truthPrices, stock and attributes come from one master repository, not parallel entries.
2. Consistent prices everywhereDatabase, page, markup and feed show strictly the same price.
3. Honest stockDisplayed availability reflects reality, updated at least daily.
4. Complete attributesDimensions, materials, compatibilities, packaging: filtering criteria are populated.

B. Structuring

5. Schema.org Product + Offer across 100% of the catalogGenerated dynamically from the database, never by hand.
6. Organization marked upIdentity, contact details and customer service readable by machines.
7. B2B fields populatedMinimum quantities, net prices, packaging (if you sell to professionals).
8. Zero errors in Search ConsoleThe structured data report is green across every template.

C. Feeds & channels

9. Complete, daily product feedMerchant Center (or equivalent) receives the whole catalog, up to date.
10. Feed ↔ site consistency verifiedA regular audit confirms feed, pages and markup tell the same story.
11. Platform compatibility identifiedYou know what your platform enables natively (UCP, ACP) and what's left to do.
12. Readable commercial termsDelivery, returns, warranties: structured HTML at stable URLs — not PDFs.

D. Agent access

13. llms.txt at the rootYour site's table of contents for AI models.
14. robots.txt open to chosen AI crawlersA deliberate policy, not an inheritance: you know who you let read.
15. An agent endpoint existsUCP profile enabled or MCP server exposed, at least read-only.
16. Customer authentication plannedThe path to per-account negotiated pricing is defined (B2B) or account linking enabled (B2C).

E. Measurement & learning

17. AI traffic isolated in analyticsVisits referred by ChatGPT, Gemini, Perplexity… are segmented.
18. Agent requests loggedYou know what agents ask for — and what they fail to find.
19. Channel costs arbitratedAgentic surface commissions are folded into your margin calculations.
20. Protocol watch in placeSomeone tracks UCP, ACP and AP2 — their specs move every quarter.

GOING FURTHER

Want a diagnosis with no blind spots? Our Commerce Agentique experts audit your site against all 20 agent-ready points — product data, structuring, feeds, agent access, measurement — and hand you a prioritised report ranked by effort-to-impact, with a concrete roadmap.

Request a full audit →

→

Is your site agent-ready? The 20-point audit