Intent matching is the process of interpreting a user’s current input and comparing the expressed goal, requirements, and supported context with available candidates to estimate relevance.
In offer discovery, the candidates are commercial offers, and the result can be a relevance score, ranking signal, or no match. Deterministic constraints and eligibility rules are evaluated separately. An intent match does not confirm availability, transaction eligibility, consent, or purchase authorization.
Intent matching in an AON query
A publisher can send natural-language text or image content through the documented intent field. AON applies supported eligibility rules and uses the supplied intent to rank eligible offers by relevance. The Query API returns structured offers or an empty offers array, not a named intent class or classifier confidence score.
See the Offer Query request modelFrom request to relevance signals
Receive the current request
The system uses the user’s words, image, or other supported input as evidence of the present commercial goal.
Identify expressed requirements
The request is interpreted for relevant meaning and explicitly stated parameters without inventing unsupported preferences.
Compare with candidate offers
Semantic signals from the request are compared with offer data after offer eligibility removes candidates that cannot be served.
Produce relevance signals
The resulting scores or ranking signals help the discovery system order candidates or return no match.
How intent matching feeds offer discovery
Intent matching supplies semantic relevance signals, while constraints, eligibility, and ranking remain distinct parts of discovery.
- 01Intent content
The user’s current text, image, or other supported input that expresses a commercial goal.
- 02Expressed requirements
The needs, preferences, and parameters stated or directly supported by the request.
- 03Eligibility rules
Remove candidates that cannot be served for the supplied context or supported constraints.
- 04Relevance and ranking
Compare the remaining offers with the intent and order them by estimated fit.
- The phrase intent matching has broader meanings in NLU, search, advertising, and recommendation systems.
- Publishers should send only context they can accurately and appropriately provide; unverified attributes should not be guessed.
- A relevance score does not override deterministic eligibility rules, merchant terms, or checkout-time verification.
How intent matching and category filtering work together
Intent matching interprets the user's goal, even when the exact product words differ. Category filtering is a separate explicit constraint: use it when results outside a known category would be invalid. The two can work together in the same offer query. See the Category taxonomy definition.
Example in practice
A five-person consultancy asks for accounting software with mobile receipt capture and simple client invoicing. The publisher sends that request through intent.content and may add a supported category constraint when the category is explicit. AON applies eligibility rules, compares the remaining offers with the stated needs, and returns ranked offers or an empty offers array.
Common questions about intent matching
How does intent matching relate to agentic offer discovery?
Intent matching estimates how well candidate offers fit the user’s expressed goal. Agentic offer discovery is the broader process that retrieves supply, applies eligibility rules, ranks candidates, returns structured offers, and handles empty results.
Must a publisher classify the user’s intent before querying AON?
No. A publisher can send supported natural-language or image content through intent.content. It may extract explicit parameters for its own workflow, but the AON Query API does not require a named intent class or classifier confidence score.
Should an AI assistant infer a missing preference?
Send only preferences or signals that are explicitly available to the integration. Optional signals and constraints can be omitted when they are unknown. When AON returns engagement refinements, use the documented request patch to narrow a later query; user-provided conditions take priority over conflicting suggested values.
When should I use intent matching instead of category filtering?
Use intent matching to interpret a broad or descriptive goal. Use a category filter when a supported category is explicitly known and results outside that category would be invalid; the filter is applied separately from the intent interpretation, so the two mechanisms can work together.
Does a relevant offer automatically mean the user is eligible?
No. Relevance estimates usefulness, while eligibility applies separate serving or transaction rules. Merchant checkout can still require authoritative verification.
Sources
- AON Offer Query APIAON documentation
- Google Dialogflow intent matchingIndustry source
- UCP Catalog SearchRelated standard