AI shopping agents are using the checkout pages retailers already built

A retailer can receive an order from an AI shopping assistant through its ordinary checkout pages. In Forter’s October network analysis, agents using browsers accounted for 65% of detected agentic orders in the final week reported, compared with 1% at the beginning of August. The dataset runs from 1 August to 7 October 2026. These are shares of AI-placed orders within Forter’s network, not shares of all online purchases.[1]

The observation also changed as Forter learnt to detect more agents reliably. Instinct and Grok Bot entered its detection record on 1 September; Muse followed on 21 September. Those dates may lag actual activity. The report, updated on 8 October, supplies no absolute order count, so the shift cannot be read as a constant-coverage measure of the whole market.[1]

The practical mechanism is straightforward. A browser agent can follow the same product pages, forms and checkout journey that a person uses. It does not need the retailer to provide a separate interface for that particular AI service. Forter describes this route as screen scraping, although completing a purchase involves acting on the website as well as reading it.[2]

That compatibility creates a useful head start: the shop already has a path to purchase. But the agent inherits the work of navigating it. A purpose-built checkout protocol gives software a different route, in which the merchant’s systems return structured information instead of requiring the assistant to interpret and operate the customer-facing pages.[2][3]

The Agentic Commerce Protocol’s seller guide makes the distinction concrete. A seller implements calls to create, retrieve, update, complete and cancel a checkout session. The seller returns prices, fulfilment options and payment capabilities; changing an address or the basket triggers an updated checkout state. The agent and the store exchange the information needed for the transaction directly.[3]

Forter is selling a way to connect these routes to a merchant’s existing operation. On 13 October it announced new capabilities including an AI activity dashboard, fraud management for agent traffic and orchestration for protocol-based checkout. Its dashboard identifies which agents are transacting, while orchestration connects the checkout journey to the merchant’s systems.[4]

The integration work depends on the store. Forter’s documentation describes connections to Shopify and Salesforce Commerce Cloud, but a custom commerce backend needs a product feed and three merchant endpoints: account, cart and checkout. The cart endpoint must return priced items, shipping options and totals. If it fails, the transaction cannot continue. A shared integration layer still needs a functioning commercial system behind it.[5]

This is why browser reach and protocol readiness answer different questions. The first determines which existing stores an assistant can attempt to use. The second gives a participating store an explicit way to describe and execute a transaction. Both routes can remain commercially useful.[2][3]

The less visible test arrives after the assistant finishes. Forter’s documentation keeps returns in the merchant’s normal process and sends orders into its backend. The purchase must remain recognisable there, with the information needed to fulfil it and handle the customer’s next request. The person may have delegated the shopping; the retailer still has an order to honour.[5]

Report an error or suggest a correction