Retailers begin blocking agent traffic at checkout

Retailers have started treating autonomous purchase attempts as a fraud signal. For the firms in our Agentic Commerce area, that turns a technical capability into a commercial problem: an agent that can complete a checkout is worth nothing if the merchant declines the transaction.

Why merchants are pushing back

From the merchant side an agent looks like the pattern they have spent a decade learning to block: rapid form completion, no mouse movement, a data-centre IP address. Fraud systems are doing what they were built to do. The deeper issue is commercial, though. Agents strip out the merchandising layer, comparison-shop on price alone, and arrive with no attribution, which removes most of what a retailer optimises for.

What this changes for the vendors

Two strategies are emerging. The first is to become an authorised partner, with a merchant-side integration and an agreed identity for the agent. That works and is slow. The second is to route purchases through a payment credential the merchant already trusts, which works until the merchant notices.

The first approach is the one worth backing. Ask any vendor in this area how many merchant integrations they hold and whether they are contractual or tolerated.

What we do not know yet

None of the three retailers has published a formal policy, so the scope of the blocking is unclear: it may apply only to unauthenticated sessions rather than all agent traffic. We have asked all three and will update this page with any response.

What to do if you are buying

If you are evaluating a shopping-agent vendor this quarter, add one item to the shortlist criteria: named merchant integrations, in writing. It is the only defence against a capability that works in the demo and fails in the field.