What Is Agentic Commerce Readiness? A 47-Point Merchant Data Checklist

In June 2026, Checkout.com's Agentic Commerce 2026 research found that 72% of UK and US merchants agree consumers will adopt agent-led shopping faster than most merchants are prepared for, while only 3% of transactions currently involve AI agents and 89% of merchants say they are actively preparing.

B2B Centr tracks the agentic commerce stack from protocol specification to merchant implementation, separating what has actually shipped from what has only been announced.

Agentic commerce readiness is now a data problem rather than a strategy problem, because agents do not browse, they parse, and anything they cannot parse they skip.

This article defines agentic commerce readiness precisely, breaks down the market conditions forcing the issue, and gives you a 47-point merchant data checklist you can audit against this quarter.

Key Takeaways

• Agentic commerce readiness means machine-readable product, price, and policy data.
• Roughly 60% of ecommerce catalogs contain gaps that cause agent exclusion.
• Feed freshness under 15 minutes is now an operational baseline, not an aspiration.
• Discovery moved to AI surfaces; most checkout stayed on merchant storefronts.
• Score your catalog across seven layers before buying any agentic commerce tooling.

The 2026 Market Reality: Demand Is Real, Conversion Is Not Yet

The demand signal is no longer ambiguous. Salesforce data shows 39% of consumers, and over half of Gen Z, already use AI for product discovery.

Morgan Stanley found 23% of Americans bought something via AI in the past month, and projects $190bn to $385bn in US ecommerce spending flowing through agents by 2030.

McKinsey puts global agentic commerce volume at $5 trillion by 2030, while Bain estimates 15% to 25% of ecommerce will move through agentic channels by the same date.

The conversion picture is far less flattering. Academic work from Kaiser and Schulze found that fewer than 0.2% of ecommerce sessions currently originate from ChatGPT referrals, and that those referrals convert 86% worse than affiliate links.

McKinsey separately measured 4.4x higher conversion for AI-generated product recommendations versus traditional search. Both findings can be true at once. The potential is real, but it only materializes for merchants whose data can support what an agent needs to do.

Checkout.com's consumer research explains part of the gap. One in four consumers say they will never delegate purchases to AI, and 27% trust no organization to operate an AI shopping agent.

Consumers would allow an agent to spend an average of £177 per purchase without further approval, below the £200 merchants assumed. Delegation starts with groceries at 41% and household supplies at 31%, not with high-consideration categories.

Why Merchant Infrastructure Became the Bottleneck

The protocol layer stabilized faster than the merchant layer.

Google's Universal Commerce Protocol launched in January 2026 with Shopify, Etsy, Wayfair, Target, and Walmart, then added cart support, catalog access, and identity linking in March 2026. At Google Marketing Live 2026, Amazon, Meta, Microsoft, Salesforce, and Stripe joined the steering group.

The Agentic Commerce Protocol, released under Apache 2.0 by Stripe, OpenAI, and Meta, remains the open discovery and catalog layer behind ChatGPT shopping. For a side-by-side comparison of how these standards interact, see our breakdown of AP2, ACP, x402, and MPP in 2026.

The most instructive event of the year was a retreat. On 24 March 2026, OpenAI stated that the initial version of Instant Checkout did not offer the flexibility it wanted, and shifted focus to product discovery while allowing merchants to run their own checkout experiences.

Walmart reportedly measured in-chat checkout converting around 3x worse than a click-through to its own site. The durable pattern that emerged is discover in AI, buy on your own site. We covered the implications in ChatGPT product discovery vs Instant Checkout.

That shift raises the stakes on data rather than lowering them. If discovery happens inside an assistant, your feed is the only thing representing you at the moment of consideration.

Mirakl research found 42% of customers abandon purchases due to insufficient product information, and estimated an average of $15m lost annually per business to poor data quality. Agents inherit those evaluation criteria and apply them at machine scale.

What Is Agentic Commerce Readiness?

Agentic commerce readiness is a merchant's measurable ability to be discovered, evaluated, transacted with, and serviced by autonomous AI agents without human intervention at any step. It is distinct from AI-assisted shopping, where a person stays in the loop, and from internal AI tooling, which helps staff but never touches an external agent.

Readiness is operational, not aspirational. A merchant is ready when an agent can complete a full loop: find the product, confirm it matches the stated requirement, verify price and availability in real time, understand shipping and returns, complete or hand off checkout cleanly, and retrieve order status afterwards.

Any break in that chain causes the agent to downgrade or exclude the merchant entirely, and agents do not send you a rejection notice.

The scale of the gap is documented. Adam Silva Consulting's compliance checks across 500 enterprise domains found 85% of respondents scoring below 40 out of 100 on their readiness index, and 72% of retailers with no UCP manifest at all. Independent estimates put roughly 60% of ecommerce catalogs in a state of missing GTINs, inconsistent attribute naming, or stale inventory states.

The 47-Point Merchant Data Checklist

Score one point per item. Audit your top 20% of SKUs by revenue first, then expand. Anything you cannot answer with evidence from your own systems scores zero.

A. Product Identity and Catalog Integrity (8 points)

1. Every SKU carries a unique, persistent product ID that is never recycled.
2. GTIN, UPC, or EAN coverage exceeds 90% of active listings.
3. MPN is populated for every product without a manufacturer GTIN.
4. Brand is set on all new products, with documented exceptions only.
5. Private-label items declare identifier_exists: false rather than leaving fields blank.
6. Feed product IDs map to checkout system IDs through merchant_item_id.
7. Variants are grouped under item_group_id instead of duplicated as separate parents.
8. Duplicate, discontinued, and orphaned SKUs are purged on a scheduled cadence.

B. Attribute Depth and Conversational Data (8 points)

9. Product category and product type map to the current Google taxonomy.
10. Color, size, material, pattern, age group, and gender are populated where applicable.
11. Dimensions, weight, and unit pricing exist for every physical good.
12. Descriptions exceed 300 characters and describe use cases rather than adjectives.
13. Compatibility fields list supported models, accessories, and replacement parts.
14. Structured Q&A pairs cover the ten most common pre-purchase questions per category.
15. Certifications, warranty terms, and sustainability claims are stored as fields, not prose.
16. Substitute and complementary SKUs are declared for out-of-stock fallbacks.

C. Imagery and Media (5 points)

17. Primary images meet the resolution floor Google enforces from 31 January 2027.
18. Every variant has its own image rather than inheriting the parent shot.
19. Lifestyle, scale, and detail shots are supplied as additional_image_link entries.
20. Alt text describes the product literally rather than repeating the title.
21. Image URLs are stable, publicly crawlable, and not blocked by a CDN rule.

D. Pricing, Promotions, and Tax (6 points)

22. Feed price matches landing page price at all times, including during sales.
23. Sale price and effective date ranges are declared, not baked into the base price.
24. Currency and country of sale are explicit for every market served.
25. Promotions are exposed through a promotions API rather than banner creative.
26. Tax and shipping settings are fully configured at account level.
27. Subscription, bundle, and quantity-break pricing are machine readable.

E. Inventory and Fulfillment Freshness (7 points)

28. Availability state updates within a 15 minute lag window at most.
29. Real-time inventory is exposed at SKU level, not category level.
30. Backorder, preorder, and availability_date are distinguished from plain out-of-stock.
31. Shipping cost, carrier, and delivery window are stated per destination.
32. Handling time is separated from transit time in the feed.
33. Store-level or regional inventory is published if you fulfill locally.
34. Feed ingestion errors trigger alerts rather than silent partial updates.

F. Structured Data and Machine Access (7 points)

35. Schema.org Product, Offer, and AggregateRating markup validates without errors.
36. On-page markup and feed data agree on price, availability, and identifiers.
37. robots.txt explicitly permits the AI crawlers and agents you intend to serve.
38. Server-side event collection captures agent traffic that never fires JavaScript.
39. Product and policy pages render server-side, not behind client-only rendering.
40. Bot management distinguishes verified shopping agents from scrapers.
41. Product endpoints respond under 200ms and checkout endpoints under 500ms.

G. Protocol, Trust, and Post-Purchase (6 points)

42. A UCP manifest is published at /.well-known/ucp if you support native checkout.
43. native_commerce is enabled on eligible SKUs.
44. Return windows, restocking fees, and exceptions are published as structured policy.
45. Review counts and ratings are exposed in a parseable format.
46. Order status, cancellation, and return initiation are available through an API.
47. Every agent-mediated transaction logs agent identity, authorization scope, and audit trail.

How to Read Your Score

• 0 to 15: Invisible. Agents cannot reliably identify or price your products. Fix sections A and E before anything else, because identity and freshness gate everything downstream.
• 16 to 30: Discoverable but not transactable. You will appear in AI recommendations inconsistently and lose the transaction at the handoff. Sections D and F are the usual failure points.
• 31 to 41: Transactable. You can support agent-mediated purchase, but post-purchase and trust gaps will suppress repeat inclusion. Close section G.
• 42 to 47: Operationally ready. Shift from remediation to monitoring, and treat drift as the ongoing risk. Our guide to making your checkout AI agent ready covers the transaction layer in depth.

What Readiness Actually Costs

Implementation timelines are shorter than most teams assume, but data cleanup is the long pole.

Shopify merchants using native protocol apps can deploy in under 48 hours. Headless setups typically take one to two weeks. Custom commerce platforms need two to four weeks for endpoint work. In every case, the dominant cost is not engineering time but cleaning titles, standardizing attributes, and fixing inventory accuracy.

Staffing is where budgets break. US market data from 2026 puts:

• PIM Manager at an average of $106,639, with a typical band of $75,000 to $132,500.
• Product Data Managers average $96,154.
• Data Product Managers average $139,710 to $147,050 depending on source, with the 75th percentile near $185,000.
• Data Governance Program Managers average $160,465.

A realistic first-year readiness team of one PIM lead, one feed engineer, and fractional governance oversight lands between $250,000 and $400,000 in fully loaded cost before tooling.

Set that against Mirakl's $15m average annual loss to poor data quality and the business case usually writes itself. It is worth noting that dual-protocol merchants have been reported to capture around 40% more agentic traffic than single-protocol stores, though that figure is vendor-sourced and should be treated as directional.

Five Mistakes That Quietly Kill Agent Visibility

• Measuring features instead of capability. Counting integrations tells you nothing. If a journey still relies on manual exception handling behind the scenes, it is not agent-ready.
• Letting feed and page data drift apart. Drift happens when product detail pages and feeds are owned by different teams. Agents cross-check, and mismatches read as unreliability.
• Treating GTINs as optional. Missing, wrong, or invented GTINs drop products out of the competitive clusters where AI surfaces aggregate listings across retailers.
• Relying on client-side tracking. Agents make direct API calls and do not execute browser JavaScript, so pixel-based measurement creates blind spots that widen as agent traffic grows. The same visibility problem shows up in citations, which we unpack in our guide to AI search visibility for B2B brands.
• Ignoring the trust layer. Consumers name spending caps at 30%, instant revocation at 29%, and easy cancellation at 28% as non-negotiables, and 75% of merchants agree real-time revocation will be critical. Our analysis of whether agentic commerce is safe and of verifiable credentials in AI agent payments covers the controls that address this.

A 90-Day Readiness Roadmap

• Days 1 to 30: Baseline. Run the 47-point audit against your top revenue decile. Pull feed error reports, validate schema markup, and measure actual inventory lag rather than the target. Document the score and the top ten failing items by revenue exposure.
• Days 31 to 60: Remediate identity and freshness. Close GTIN and MPN gaps, fix variant grouping, rewrite titles for the top 20% of SKUs, and get inventory sync inside 15 minutes. Exhaust free platform automation before evaluating paid enrichment tooling.
• Days 61 to 90: Open the machine access layer. Update robots.txt and bot rules, deploy server-side collection, publish structured return and shipping policy, and stand up protocol endpoints if your category justifies native checkout. Then rescore and set a monthly drift check.

Readiness is not a launch. It is a maintained state, and the merchants who treat it as ongoing operations rather than a project are the ones whose scores hold.

Conclusion

B2B Centr covers the agentic commerce stack at implementation depth, tracking what protocols require of merchants rather than what vendors claim they deliver.

Agentic commerce readiness is the measurable ability of your catalog, endpoints, and policies to be parsed and acted on by autonomous agents.

The 2026 evidence is consistent: consumer demand is forming faster than merchant infrastructure, discovery has moved to AI surfaces while checkout largely stayed on merchant sites, and the merchants capturing value are the ones with complete identifiers, sub-15-minute inventory sync, machine-readable policy, and endpoints agents can trust.

Score yourself across the seven layers, fix identity and freshness first, and treat the result as an operational metric you monitor monthly.

Read Next

• 2026 Payroll Spending Report: How Much Do Companies Spend on Payroll Per Year?
• ChatGPT Product Discovery vs Instant Checkout: What Merchants Need to Do Differently Now
• What Are Verifiable Credentials in AI Agent Payments? W3C Standards Explained

FAQs

1. What is agentic commerce readiness?

Agentic commerce readiness is a merchant's measurable ability to be discovered, evaluated, transacted with, and serviced by autonomous AI agents without human intervention. It spans product identity, attribute depth, media, pricing, inventory freshness, structured data, machine access, and post-purchase APIs.

2. How many merchants are ready for agentic commerce in 2026?

Very few merchants are ready for agentic commerce in 2026. Compliance testing across 500 enterprise domains found 85% scoring below 40 out of 100 on readiness, and 72% of retailers had no UCP manifest published at all.

3. What is the most important data field for agentic commerce readiness?

The most important data field for agentic commerce readiness is the product identifier, specifically GTIN. Missing or invalid GTINs remove products from the competitive clusters AI surfaces use to aggregate listings across retailers, making everything downstream irrelevant.

4. How often should product feeds update for AI shopping agents?

Product feeds should update for AI shopping agents within a 15 minute lag window at most. Availability and pricing that lag beyond this cause failed transactions, which agents record as unreliability and use to deprioritize the merchant.

5. Does agentic commerce readiness still matter after Instant Checkout was pulled back?

Agentic commerce readiness matters more after Instant Checkout was pulled back, not less. OpenAI shifted to product discovery in March 2026, which means your feed is now the primary signal determining whether you appear in AI shopping results at all.

Disclaimer: This content is provided for informational purposes only and does not constitute legal, financial, or compliance advice. Protocol versions, governance arrangements, and partner counts cited here reflect publicly announced milestones as of August 2026 and are moving quickly. Adoption figures come from vendor and foundation announcements with differing methodologies and should be treated as directional signals rather than guaranteed outcomes.