On this page
Agentic offer discovery is the process through which an AI app finds and evaluates commercial offers for a user's request before sending them to a merchant.
The app or offer system retrieves candidates, checks which can be shown, and compares their relevance. In AON, results include the offer's supported next action and tracking context, not just product facts. The app decides what to present; the merchant confirms current price, availability, and purchase terms at checkout.
Multi-merchant discovery example for an AI assistant
A traveler asks for a compact camera within a budget. The assistant can use product data to compare size and specifications, then query an offer-discovery service for commercial options from connected merchants. It evaluates the returned options and can show a useful merchant link to the traveler.
The assistant still needs to check whether each option meets the request. A budget written in the user's message is not automatically a strict API filter. If a requirement remains unverified, the assistant should explain that before the user continues to the merchant.
How multi-merchant offer discovery works
Multi-merchant discovery lets an AI app compare commercial options from more than one connected source. The integration searches its approved, active supply, not necessarily every retailer.
Capture the request
The app sends what the user is looking for, along with accurate context and any filters the integration supports. It should not invent missing requirements.
Find candidate offers
The offer system searches connected sources for commercial options that could fit the request.
Check which offers can be shown
The system removes offers that fail its serving rules for the request. A relevant product can still be excluded if it is not eligible.
Compare relevance and rank
The system orders eligible offers by how useful they appear for the request and the place where they would be shown.
Return offers or an empty result
The app receives structured offer records, or no offers when nothing should be returned.
Choose whether to present a commercial option
The team operating the app decides which returned offers to show, explains the commercial relationship, and keeps the supported merchant link or action. It does not have to recommend every result.
Product research vs offer discovery vs checkout
These stages answer different questions and should not be described as one universal purchasing API.
| Stage | Question it answers | Result and boundary |
|---|---|---|
| Product research | Which items have the required characteristics? | Product APIs and feeds provide facts and seller observations; they do not automatically establish commercial permission. |
| Offer discovery | Which available commercial options can fit this request? | Offer records include a supported next action and context. A returned offer is not a reserved order. |
| Checkout | Can the merchant accept this purchase now? | The merchant confirms current price, availability, eligibility, payment, and order acceptance. |
Where do offer-discovery results come from?
An offer-discovery service needs connected merchants or other authorized supply providers. Their available products, countries, categories, and permissions determine what the service can search.
Supply integration prepares those sources before a user searches; it is not another checkout step. In AON, a supply Partner maps permitted upstream information into the Offer format and maintains outcome reporting.
What to do when no eligible Offer is found
An empty result is valid when no active eligible supply matches the request. Do not turn it into an unrelated recommendation or infer that the entire market lacks a suitable product.
The app can continue with product research, clarify the request, or adjust supported filters when appropriate. It should explain any change rather than silently dropping a requirement. Keep the request ID for technical diagnostics.
Does offer discovery confirm price and stock?
No. Discovery reflects available source information at that stage. The merchant or checkout system confirms current price, stock, and binding terms.
An offer can look relevant but fail the rules for being shown. A high ranking also does not authorize the assistant to buy it for the user.
Offer discovery in an AON query
An AI app sends AON the user's request, as text or an image, along with supported context and filters. AON checks which offers can be returned and ranks eligible options by relevance. The app decides whether to show them and how to explain each option.
AON retrieves supply from approved Partners, the businesses responsible for providing authorized offers and reporting referral outcomes. A request can validly return no offers. The app should not invent a result or treat a returned offer as confirmation of an order.
See the Offer Query APICommon questions about Agentic offer discovery
Is intent matching the same as agentic offer discovery?
No. Intent matching measures how well the expressed need fits candidate offers. Agentic offer discovery is the broader process that retrieves supply, applies eligibility rules, ranks candidates, returns structured offers, and handles empty results.
What happens when no offer matches an AON query?
AON can return a successful response with an empty offers array when no offer should be returned for the request. This can reflect no suitable eligible supply or a runtime rule that suppresses an offer. An empty result is valid, not an API failure or a reason to substitute a weak or ineligible offer.
What should an AI product do after AON returns no offers?
Continue helping the user without forcing an unrelated commercial offer. Clarify the request or adjust a supported filter when appropriate, without silently dropping the user's requirements. For technical diagnostics, keep data.request_id and follow the documented Query guidance.
Does an AON offer-discovery response confirm final price and availability?
No. It reflects the information available during discovery. The merchant or checkout system must confirm current price, availability, eligibility, and binding terms.
Sources
- AON Offer Query APIAON documentation
- UCP Catalog capabilityRelated standard
- OpenAI product discovery in ChatGPTIndustry source
- Schema.org OfferRelated standard