AI Shopping and Agentic Commerce Readiness
MERCHANT READINESS ASSESSMENT
Score the operating system, not the AI marketing language
An SME is ready to test AI shopping when it can publish accurate product and offer data, honour the promise at checkout, identify the channel, resolve exceptions and correct errors. A badge, plugin or new file does not replace those controls.
01 – PRODUCT
Prove catalogue truth
Verify identifiers, variants, attributes, images, price and availability for a commercially meaningful sample.
02 – TRANSACTION
Protect the promise
Revalidate item, stock, address, delivery, promotion, payment and policy at merchant checkout.
03 – OPERATIONS
Own the channel
Monitor feed errors, referrals, incorrect representations, cancellations, returns and customer-support exceptions.
Score one product family before integrating a catalogue
Complete the checklist with evidence, close red items and run a customer journey from conversational discovery to merchant fulfilment.
AI shopping readiness is an evidence-based release decision. It asks whether the merchant’s product data and operations can support a new discovery surface without misleading the customer. The assessment should be completed for a defined catalogue and channel. A store may be ready for one well-managed product family and unready for thousands of incomplete variants.
Read the agentic-commerce comparison for the model decision, the structured-data guide for implementation, and the ecommerce service when the store, feed and checkout need one controlled source.
How to use the score
Rate each control zero when no accountable evidence exists, one when it is partly implemented or manually dependent, and two when it is defined, tested and monitored. A high total cannot cancel a critical zero in identity, price, availability, checkout or policy. Record the reviewer, date, sample and evidence link. Reassess after material catalogue or provider changes.
| Readiness level | Interpretation | Action |
|---|---|---|
| Hold | Critical product or transaction controls are missing | Do not integrate the affected catalogue |
| Limited pilot | Core controls pass for a narrow sample | Test one family with human monitoring |
| Controlled expansion | Publishing and exception processes are repeatable | Add categories in reviewed batches |
| Operational | Quality, correction and outcome monitoring are sustained | Consider deeper permissions cautiously |
1. Product identity and variants
- Every purchasable variant has one stable internal identity.
- GTIN, MPN, brand and SKU are used accurately where applicable.
- Size, colour, material, dimensions and compatibility attach to the correct variant.
- Included items and optional accessories are distinguishable.
- The primary image represents the selected product and condition.
- Discontinued and replacement relationships are governed.
2. Offer and availability truth
Check price, currency, tax presentation, condition, availability, delivery promise, promotion and seller. Define the stock state and freshness threshold used by the channel. OpenAI’s March 2026 product-discovery update describes merchant product feeds and promotions through ACP and merchant-owned checkout as the current direction. Provider capabilities can change, so review current terms and specifications at implementation time.
3. Visible pages and machine-readable outputs
| Control | Evidence |
|---|---|
| Visible product page | Customer can understand and buy the exact variant |
| Product structured data | Current feature requirements pass and match the page |
| Merchant feed | Items are accepted, current and mapped to controlled fields |
| Canonical URL | The preferred product or variant destination resolves correctly |
| Policy access | Shipping, returns and contact routes are reachable |
Google explains that structured data and Merchant Center feeds can both supply product information, but eligibility does not guarantee display. Markup must match visible content. OpenAI’s merchant-feed terms require merchants to be responsible for submitted data under the applicable agreement. Treat feeds as published business records, not experimental copy.
4. Checkout and confirmation
At checkout, revalidate product, variant, quantity, price, currency, promotion, address, delivery method and availability. Show the merchant identity and applicable terms. Do not assume a discovery platform’s earlier product representation is still current. Record channel attribution only when technically reliable and never weaken fraud or payment controls for a new source.
5. Fulfilment, returns and support
- Order source and external reference are retained.
- Warehouse or supplier receives the exact variant and service promise.
- Shipment and tracking return to the customer journey.
- Cancellation, partial fulfilment and substitution rules are explicit.
- Returns connect to the original order line and payment event.
- Support can see what the customer was shown and what checkout confirmed.
6. Monitoring and correction
Track rejected products, stale offers, price mismatches, incorrect variants, out-of-stock referrals, checkout failures, cancellations, returns and customer corrections. Assign an owner and response target. A channel should be suspendable without disabling the core store. Record when a correction reaches the page, feed and downstream system.
7. Review provider and channel change
Record the provider, integration version, catalogue scope and terms reviewed. Re-run critical checkout and feed tests when a platform changes its product specification, discovery experience, merchant agreement or regional availability. A capability announced for one partner or market should not be assumed available to every SME. Keep the pilot reversible so the team can pause a channel while correcting data or policy without disabling the merchant store.
Evidence pack and release gate
The release pack should contain the scored checklist, sample products, field mappings, validation results, provider terms reviewed, checkout test, exception test, responsible owners and rollback route. Release only the exact products that passed. A future platform announcement is not a reason to bypass the gate.
Worked assessment
A homeware store scores 200 products. Product identity and checkout are strong, but sixty variants share generic images and supplier availability updates only once a week. Instead of releasing the whole catalogue, it chooses twenty products with unique images and retailer-held stock. The team validates pages, Product markup, a feed, checkout, fulfilment and returns. Feed errors and incorrect representations have a named queue. The remaining products stay outside the pilot until image and availability evidence improves.
Sources checked
- OpenAI: Powering Product Discovery in ChatGPT
- OpenAI: Merchant Feed Terms of Service
- Google Search Central: Product structured data
- Google Merchant Center: Product data specification
Reviewed by
Mitrend Digital editorial team
2026-07-17
Evidence used for this page
Reviewed against OpenAI’s March 2026 product-discovery update, current merchant-feed terms, Google Product structured-data guidance and Merchant Center specifications. Includes an original scored readiness checklist.
Turn the guide into a practical next step
This resource provides general implementation guidance. Verify platform settings, tax, legal, payment and operational requirements against the current business context before making a live change.