Agent-Readiness Checklist

Agents do not read design, they read data. This list measures how ready a business is for AI agents across 12 dimensions and 148 auditable checks.

Agent-readiness checklist parts

Agentic Readiness

Progress 0 / 148 completed
Part 1 · Discoverability and agent access
0/16
High
Medium
High
Medium
High
High
Medium
Low
High
Medium
Medium
Low
Medium
High
High
High
Part 2 · Machine readability and structured data
0/18
High
High
High
Medium
Medium
Medium
Low
Low
High
High
Medium
High
Medium
Low
Low
High
Medium
Low
Part 3 · Product data and feed readiness
0/20
High
High
High
High
High
Medium
Medium
Medium
Low
High
High
High
High
High
Medium
High
High
Medium
Medium
Low
Part 4 · Protocol and integration layer
0/13
Medium
High
Medium
Medium
High
Medium
Low
Medium
High
High
Medium
Medium
Medium
Part 5 · Payments, commercial terms and unit economics
0/10
High
High
High
Medium
High
Medium
Medium
High
High
Medium
Part 6 · Operations, inventory and service reliability
0/10
High
High
High
High
Medium
Medium
High
Medium
Medium
Medium
Part 7 · Trust, identity and governance
0/10
Medium
High
High
Medium
High
High
Medium
Medium
High
Medium
Part 8 · AI visibility (GEO and AEO)
0/13
High
High
Medium
Medium
Medium
Medium
Medium
High
Medium
Medium
High
Medium
Low
Part 9 · Measurement and attribution
0/10
High
High
High
Medium
Medium
High
Medium
Medium
High
Medium
Part 10 · AI advertising (GEA) readiness
0/10
High
High
High
Medium
Medium
High
High
Medium
Medium
Low
Part 11 · Legal, compliance and data
0/10
High
High
High
Medium
High
High
High
Medium
Medium
Medium
Part 12 · Organisation, capability and sourcing
0/8
High
High
Medium
High
Medium
Medium
Medium
Medium

The chain rule: order matters more than effort

This is not a grab bag of improvements to work through in any order. The chain is:

Access → Readability → Data → Visibility → Protocol → Transaction → Measurement

Skip a link and the spend above it is wasted. A site with a flawless feed that blocks crawlers is never found in the first place. A store that appears in AI answers but carries stale stock data loses the agent’s trust in a way that is slow to recover.

Scope: the checks are split so they work for product businesses and for service and B2B businesses. A B2B firm with no product feed can skip Parts 3 and 5. Parts 1, 2, 7, 8 and 9 apply to everyone.

Scoring and maturity tiers

Critical items are worth 3 points, important items 2, maturing items 1. Drop any part that does not apply to you out of the denominator and score as a percentage. The full list is worth 354 points across 70 critical, 66 important and 12 maturing items.

PercentageTierWhat it meansNext move
0-25%Tier 0 · InvisibleThe agent cannot read the site or cannot understand the productParts 1 and 2
26-50%Tier 1 · AccessibleThe agent can read it, but the data is not trustworthyParts 3 and 6
51-75%Tier 2 · Feed-ReadyThe data is trustworthy, but there is no visibility or measurementParts 8 and 9
76-90%Tier 3 · Agent-NativeThe channel is live and measuredParts 4 and 10
91%+Tier 3+ · ReferenceMature, governance and trust layers includedLead a pilot

The rule that overrides the score: you have not reached a tier until every critical item in that part is met, however high the percentage. If OAI-SearchBot is blocked in robots.txt, a score of 80% still means Tier 0.

A 0-30-60-90 day sequence

Days 0-30 · Access and readability. All of Part 1, then 2.1 to 2.12, then 9.1 to 9.3 and 9.9 for the baseline, then 8.1 and 8.2 for the prompt set and first measurement. Outcome: a site open to crawlers with validated schema, and a measurement baseline.

Days 30-60 · Data and reliability. All of Part 3, then 6.1 to 6.4 and 6.7, then 5.1 to 5.3 for unit economics, then 12.1 and 12.2. Outcome: a validated feed, 98% stock accuracy, and a decision on which SKUs belong in the channel.

Days 60-90 · Channel, visibility and governance. Items 4.1, 4.4, 4.5 and 4.9, the rest of Part 8, then 7.3, 7.5, 7.6 and 7.9, then Part 10 if you are targeting export markets, and a review of Part 11. Outcome: live products in at least one agentic channel, a measured visibility change, and a written authority and rollback policy.

Red flags

If any one of these is true, stop everything else:

  1. The WAF or robots.txt blocks legitimate search and agent crawlers.
  2. Price or stock differs between the feed and the site.
  3. Product information renders only through JavaScript.
  4. The feed contains invented GTINs. A fabricated barcode is a permanent account risk.
  5. Reviews or AggregateRating are fabricated.
  6. Stock accuracy is below 90%.
  7. No logs are kept for agent transactions.
  8. Commission plus processing fees exceed product margin, and the channel is open anyway.
  9. Budget was allocated for the domestic market without knowing ChatGPT Ads does not target Turkiye.
  10. The vendor contract is billed by the hour with no verifiable milestone.

Honest limits

What should be said out loud when this list is presented:

  • It does not guarantee sales. Agentic commerce today is mostly a discovery and visibility channel; the first wave of embedded checkout was pulled back for behavioural reasons. Most of the investment belongs in feed quality and AI visibility, not checkout integration.
  • The Turkiye asymmetry is real. Part 10 and some of Part 4 currently return nothing in the domestic market. The value is in export markets.
  • Measurement is incomplete. A meaningful share of AI traffic cannot be captured by any setup. Treat the numbers as a floor.
  • The protocols move. ACP, UCP and AP2 specifications change quarterly. This list reflects August 2026; the checks will hold, the technical detail will not.
  • Be sceptical of industry statistics. Claims like “AI traffic converts X times better” are usually single-source and selectively sampled. Your own baseline is the only evidence that counts.

Technical specifications should be re-verified quarterly against primary sources: developers.openai.com/commerce, agenticcommerce.dev, the Universal Commerce Protocol repository and ap2-protocol.org.

Back to top