AI Agent initiated orders - API

I am exploring the UCP (Universal Commerce Protocol) led sales and how its data will look like when queried via GraphQL. What will be the value of UTMParameters source for UCP sale initiated by AI agent. Will it be something like AI-Agent or Chat-Agent or Gemmini-Agent .

The idea is how do we differentiate order if its AI enabled sale or a classic web-based.

Ref: UTMParameters - GraphQL Admin

Any insight or suggestions ?

I did further digging - Could be we could query Orders:
Orders.ChanelInformation.ChanelDetail to distinguish between AI Agent-originated order or its human originated order ? Not sure if we have ENUM value there?

Hi again @Gagan_Arora

From my understanding you’ll be able to differentiate orders from specific agents/platforms as they’ll be treated as unique sales channels, which should also be query-able via API. There’s a bit more info on this from our merchant help docs.

Hey Liam - Thank you for the insight. Quick followup:

I want to clarify one implementation detail to ensure we model this correctly on our side of execution.

In the Admin GraphQL API, is the order’s sales channel attribution exposed as a stable, enum-like value (for example via channelInformation.channelDefinition.handle or a similar canonical field), or is it only a human-readable label?

Specifically:

  1. Is Online Store / Web (Classic Human-initiated orders) represented by a single constant value across all orders placed via the standard storefront checkout? If yes then what would be that constant value ?

  2. Will each agentic storefront (e.g. ChatGPT, Copilot, etc.) have its own distinct, stable channel identifier/handle that is guaranteed not to overlap with web?

  3. Can we safely assume that any order whose channel identifier ≠ contant value (from point 1) originated from a non-traditional (agentic or marketplace) checkout flow?

We want to confirm whether this field is safe to treat as an enum for filtering, analytics, and automation logic, rather than relying on display names.