Two new protocols are getting a lot of attention in e-commerce right now: OpenAI’s Agentic Commerce Protocol (ACP) and Google’s Universal Commerce Protocol (UCP). Both let an AI agent complete a checkout without the shopper ever leaving the chat window, and both are already live with major platforms.
For affiliates, neither protocol provides a natural attribution path. ACP and UCP operate at the merchant and platform level. When Shopify tags an order as ChatGPT-referred, that happens because Shopify and OpenAI have a direct technical integration. An affiliate promoting the same product through a network has no access point into that relationship.
The attribution problem extends to merchants themselves:
A tracking method developed more than a century ago, before cookies, pixels, or even the internet existed.
In his 1923 book Scientific Advertising, Claude Hopkins formalized the idea of a keyed response code: assign a distinct code to each publication or mailer, then use the code returned with a customer’s order to identify what drove the sale.
It’s the same idea behind promo-code attribution today, which affiliate platforms already use for sales that happen without a trackable click.
This approach solves a key piece of the agentic attribution problem: give the agent an identifier it can carry from the recommendation to the checkout.
We put together a full breakdown covering:
For affiliates, neither protocol provides a natural attribution path. ACP and UCP operate at the merchant and platform level. When Shopify tags an order as ChatGPT-referred, that happens because Shopify and OpenAI have a direct technical integration. An affiliate promoting the same product through a network has no access point into that relationship.
The attribution problem extends to merchants themselves:
- GA4 can't register an agentic purchase, since the transaction happens through an API call instead of a browser session, with no page load and no thank-you page to fire a pixel from
- Paid ChatGPT accounts and Gemini's Deep Research mode strip out referrer data entirely
A tracking method developed more than a century ago, before cookies, pixels, or even the internet existed.
In his 1923 book Scientific Advertising, Claude Hopkins formalized the idea of a keyed response code: assign a distinct code to each publication or mailer, then use the code returned with a customer’s order to identify what drove the sale.
It’s the same idea behind promo-code attribution today, which affiliate platforms already use for sales that happen without a trackable click.
This approach solves a key piece of the agentic attribution problem: give the agent an identifier it can carry from the recommendation to the checkout.
We put together a full breakdown covering:
- How to get a code assigned through your network
- Which e-commerce offers work best with promo codes
- Where to place the code so an agent is likely to carry it forward
- What to watch for once it’s live, including leaked codes and click-versus-code conflicts that can quietly distort your numbers
