How to Attribute Orders From AI Agents

From Referral Traffic to Real Transactions

More customers are using AI services to discover products and businesses. But merchants still face a basic measurement problem:

How can they tell whether an order came from general AI traffic or was directly completed by an AI agent?

TL;DR

• AI traffic does not automatically equal an AI-driven order.
• A customer's path can become difficult to track when they move between an AI service and a merchant website.
• Merchants need to distinguish AI exposure, AI referrals, and agent-handled transactions.
• Agentic Store can record transaction paths handled directly through DeepLumen.
• The key is to measure transaction events, not only traffic sources.

AI Traffic Is Not the Same as an AI Order

When a customer visits a merchant website through an AI service, the merchant may be able to identify the referral source.

But a visit does not necessarily lead to a purchase.

The customer may:

• Leave the website without buying
• Search for another product
• Compare several options
• Switch devices
• Return later through a different channel
• Complete the order outside the original session

This means a merchant may know that someone arrived from an AI service without knowing whether that AI interaction directly influenced the final order.

As Agentic Commerce develops, this distinction becomes increasingly important.

Three Different Levels of AI Activity

Merchants can start by separating AI activity into three levels.

AI Exposure

An AI system mentions a merchant, product, or service to a customer.

AI Referral Traffic

A customer clicks from an AI service and visits the merchant's website or another merchant-owned channel.

Agent-Handled Transaction

An AI agent understands the customer's request, matches it with a product or service, and submits an order through a connected merchant system.

These outcomes are different and should not be measured with the same metric.

Why Traditional Attribution Becomes More Difficult

Traditional ecommerce analytics often follows a path like this:

Ad or search -> Website visit -> Product view -> Add to cart -> Checkout -> Order

An AI agent may complete several steps before the customer ever reaches a website.

For example, the agent may:

• Understand the customer's request
• Compare available products
• Read a menu
• Check availability
• Select an item
• Submit an order

If these activities take place between an AI service and a merchant system, traditional web analytics may not capture the full path.

The merchant may see the final order without seeing how the order was created.

Recording the Direct Transaction Path

Agentic Store connects AI agents with existing merchant systems and can record transaction paths handled directly through DeepLumen.

This gives merchants a way to look beyond basic referral data and understand more of the transaction process, including:

• What type of request the agent received
• How the agent matched the request to a product or menu item
• Whether availability or business rules were checked
• Whether the order was submitted to the merchant system
• Whether the merchant accepted and fulfilled the order

This helps merchants distinguish transactions handled through DeepLumen from broader organic AI traffic that may be harder to verify.

A Practical Measurement Framework

A useful Agentic Commerce measurement framework can include five stages.

1. Request

What did the customer ask for, and where did the request originate?

2. Matching

Did the AI agent find a product, menu item, or service that matched the request?

3. Validation

Before submitting the order, did the agent confirm availability, current menu information, or relevant merchant rules?

4. Submission

Was an order actually submitted to the merchant system, or did the interaction stop at a recommendation?

5. Fulfillment

Did the merchant receive and process the order?

Depending on the merchant's systems, additional statuses may include accepted, completed, canceled, or unavailable.

The exact data available will depend on the merchant's systems and integration setup. The important principle is to distinguish a recommendation from an actual transaction event.

Why Attribution Matters

Without a clear view of AI-driven transactions, merchants may struggle to answer important questions:

• Is AI traffic generating revenue?
• Which products are most likely to be selected by AI agents?
• Which customer requests are best handled conversationally?
• How do agent-handled transactions compare with website purchases?
• Should the merchant invest more in Agentic Commerce?

Attribution is not only about reporting results. It can also help merchants improve their product information, menu structure, inventory updates, and transaction workflows.

Attribution Should Not Overstate Performance

Not every AI mention should be counted as an order.

Not every AI referral should be treated as a completed purchase.

A more reliable approach is to separate:

• Merchant or product exposure
• AI-referred visits
• Agent-assisted product matching
• Agent-submitted orders
• Completed fulfillment

This gives merchants a clearer view of where AI is contributing to the customer journey.

FAQ

Conclusion

AI Shopping is changing how customers discover products, and Agentic Commerce is changing how merchants think about attribution.

In the past, merchants focused on ad clicks, website visits, and conversion rates. As AI agents become more capable, merchants will also need to understand whether an agent directly participated in product selection and order execution.

By connecting AI agents with merchant systems and recording direct transaction paths, DeepLumen is helping create a clearer way to measure the business impact of Agentic Commerce.

Learn how DeepLumen helps merchants understand and measure agent-driven transactions.