Imagine your regular customer asks an AI shopping agent to reorder a product. Another store charges less than yours, with the same delivery and return conditions.
The customer has a private coupon in their member account that would earn 500 points. Assume those points can be applied as a discount toward their next purchase. Your offer might suit them better, provided they expect to use that reward.
The coupon needs to be activated before they can earn those points. Can the agent find it, activate it, and explain the tradeoff? The comparison depends on benefits held inside the member account, beyond the public product price.
TL;DR
Loyalty can change which offer makes sense. A purchase earning usable rewards may suit a customer better than a cheaper alternative. Shopping agents need redemption conditions and an understanding of the customer’s preferences to judge that trade-off.
UCP can carry calculated discounts and expected points. A merchant can apply percentage, bundle, quantity, or points multiplier rules and return the outcome. Supporting the calculation does not automatically make the underlying offer discoverable.
Private coupons expose gaps in the standard interfaces. The discount and loyalty extensions reviewed do not define a complete way to list a member’s coupons, activate them persistently, or compare their qualification and stacking rules.
UCP can carry the result of your loyalty rules
The Universal Commerce Protocol (UCP) gives commerce systems and shopping platforms a shared way to exchange information and perform actions. The references in this article use the 2026-08-25 specification.
UCP’s discount extension accepts known codes through cart or checkout requests. This partial request uses the codes field defined in its JSON schema:
{ "discounts": { "codes": ["MEMBER10"] }}
That list replaces previously submitted codes, so include any code you want to preserve. An empty list clears the session codes. The standard does not define account coupon deactivation through that operation.
The merchant calculates the discount and returns the result. In the response, discounts.applied describes the reductions, allocations identifies where they apply, and priority gives their calculation order.
The loyalty extension in UCP can also return expected points. One reward entry could look like this:
{ "currency": { "name": "Club points", "code": "PTS" }, "earning_forecast": { "amount": 500, "breakdown": [ { "id": "base", "amount": 100, "description": "Base points" }, { "id": "bonus", "amount": 400, "description": "Bonus from 5x offer" } ] }}
This illustrative entry belongs inside a membership’s rewards array. The merchant can apply a 5x points rule and report the result without a dedicated multiplier field. The integration still needs to define how a points coupon is selected and activated.
Check your loyalty program against these 16 scenarios
The question is whether UCP can support all scenarios of your loyalty program within its standard scope. Here is how I would check a loyalty program against those extensions.
Basket offers
| Scenario | What UCP can carry | What needs an additional interface or model |
|---|---|---|
| Fixed discount | A known code and calculated reduction. | Discovery of the private code. |
| Percentage discount | The resulting monetary reduction. | A prospective percentage rule for advance comparison. |
| Minimum spend | The result after the merchant checks the threshold. | Qualification terms, including the relevant price basis. |
| Bundle or mixed products | The resulting basket and allocated discounts. | Which combinations and variants qualify. |
| Quantity thresholds | Recalculated outcomes for the submitted quantity. | A discoverable schedule for two, three, or five items. |
| Free gift | A basket containing the resulting items. | Gift qualification and selection. |
| Points multiplier | Total forecast and base or bonus contributions. | A typed multiplier rule and coupon activation behavior. |
A merchant can correctly quote a bundle discount even if the agent has not received a structured rule explaining which extra product would unlock it. Descriptions can explain conditions; a shared rule model would let different integrations interpret them consistently.
Member accounts and reward lifecycle
| Scenario | Existing foundation | What remains to define |
|---|---|---|
| Private coupon wallet | Authenticated evaluation of a known code. | Listing the member’s available coupons. |
| Persistent activation | Replace or clear session codes. | Account activation and deactivation operations. |
| Exclusive or stacked offers | Applied priority and combination rejection messages. | Advance discovery of compatible choices. |
| Channel, country, or variant limits | The merchant evaluates its rules. | Portable coupon conditions for advance checks. |
| Single use and redemption caps | The merchant can reject an unavailable code. | Coupon reservation, remaining uses, and release behavior. |
| Pending or expiring points | Expected earnings. | Availability, expiry, and spendable balances. |
| Points redemption | A redemption discount reflected in totals. | Operations for spending points and explaining their value. |
| Progress toward a higher tier | Current active tier information. | Qualification progress and future benefit comparisons. |
| Partial returns | General order adjustments. | A shared way to report points removed or coupons restored. |
These distinctions follow from the loyalty and discount interfaces, together with UCP’s broader order adjustment model.
The member account needs a coupon interface
Eligibility, activation, and redemption are different facts. A coupon can belong to the customer and still qualify for the basket while remaining inactive.
I would separate reading available benefits, previewing their effects, and changing account state. A preview should make clear whether it changes anything. Comparing two offers should not unexpectedly consume a coupon or leave a different offer active in the customer’s app.
For exclusive coupons, expose the permitted alternatives before the agent changes the selection.
Google’s 15 October 2026 Universal Cart launch describes promo codes and payment perks and includes Walmart among checkout merchants. It does not establish a complete private coupon workflow.
Google’s UCP loyalty guide currently limits verified loyalty benefits to checkout and describes personalized search as future work. Protocol coverage and platform availability need to be checked separately.
More points can justify a higher price
Let’s look again at the earlier shopping example. If 500 points provide a discount that the customer will use on a later purchase, paying a bit more today may make sense.
That judgment changes if the points expire before the customer returns, need a larger balance to redeem, or exclude the products they normally buy. A reward forecast alone cannot settle the comparison.
The UCP loyalty model describes reward units and precision. Its decimal_places field specifies precision, not conversion into euros. The loyalty extension explicitly leaves redeemable balances and stored value to a negotiated payment instrument or a future redemption capability.
However, we must also take into account that we’re just optimizing for specific cases. Two customers can give very different instructions about how to handle loyalty points:
Always prefer stores where I am participating in the loyalty program, even if prices are up to 3% higher.
You may spend more if the rewards I expect to redeem outweigh the difference.
You may spend more if the cash total is lower because you can apply any points value from my account that would expire within the next 2 months.
The shopper’s agent will make these decisions. On the side of the store’s agentic commerce implementation, what matters most is to factor in all these scenarios and optimize for them.
A coupon must identify the exact item
It is common across WooCommerce brands to sell distinctive, carefully crafted goods. These may not always come with assigned GTINs. Their products still need reliable identifiers when a coupon applies to a specific item.
A discount for the blue bowl must not accidentally qualify the green one. Keeping explicit mappings between product or variation IDs, the identifiers emitted by the feed connector, coupon targets, and transaction lines. Product identity, in the form of identifiers within structured data, is less relevant when just selling on your online shop. Shopping agents connect from externally. Clean structured data helps avoid mistakes.
Google’s product ID documentation calls for a stable id and recommends a suitable SKU. That merchant identifier is separate from an assigned GTIN.
Follow Google’s identifier rules for products without qualifying assigned identifiers. Keep genuine identifiers where they exist, and never invent a GTIN or guess an MPN if you don’t have it.
For public discovery, Google’s loyalty markup defines programs under Organization. Product member prices reference their tiers through validForMemberTier. Check that markup with the Rich Results Test. Merchant Center’s program settings take precedence when both are supplied. These describe published benefits; private coupon codes and member state belong in the account interface.
Keep pricing in the commerce system
WooCommerce offers Points and Rewards for earning points and redeeming them for discounts. Merchants can configure conversion rates and redemption limits. Those rules provide a loyalty foundation; an agent integration still needs an interface to use them.
For an integration, I would start with the installed extension’s documented integration points and the store’s existing cart calculations. Calculating discounts again from campaign descriptions would create another version of the rules to maintain.
An adapter connects the agent to those systems. I would give it five responsibilities:
| Responsibility | Required behavior |
|---|---|
| Discover | Return benefits accessible to this authenticated member, with eligibility and activation state. |
| Preview | Return accurate totals, reward forecasts, and rejected combinations with defined side effects. |
| Activate | Change the authorized selection, with explicit retry and conflict behavior. |
| Map | Translate the confirmed quote into supported UCP totals, discounts, and loyalty data. |
| Reconcile | Update coupon usage and reward state after purchase, cancellation, or refund. |
These are proposed integration responsibilities, not standard UCP endpoint names. Each operation needs permissions tied to the customer the agent represents. A preview also needs to say how long its result is valid and what must be checked again before purchase.
UCP already defines checkout expiry, retry protection, and order adjustments. The adapter must connect those controls to the loyalty system’s state. A checkout expiry timestamp alone does not specify when a coupon reservation must be released or a points hold restored.
Six checks before calling the integration ready
When you integrate your loyalty program when making your store ready for agentic commerce, I’d run these tests:
- Private discovery. The eligible member can retrieve an inactive coupon. Another member cannot retrieve it.
- Preview behavior. Comparing alternatives preserves persistent state when the interface promises a preview without changes.
- Price and reward parity. Fixed, percentage, bundle, quantity, and multiplier cases match normal checkout. Include a product without a GTIN and verify the exact variant.
- Competing sessions. Another session using a single-use coupon produces the defined conflict or rejection.
- Retries. Repeating an activation or checkout request does not duplicate the corresponding coupon or points transaction.
- Returns and CRM. A partial refund follows the program’s rules, and later emails reflect the resulting coupon and points state.
UCP provides a useful foundation for loyalty, but the extensions reviewed here do not cover the complete workflow. Your loyalty system may include scenarios that are not easily understood by shopping agents. Your online shop system can calculate complex discounts and report expected points. Can it also make them effectively available to agents?

Leave a Reply