Agentic Commerce
Definition
Agentic commerce describes a purchase journey in which an AI agent, acting on a person's instructions and within limits the person sets, searches for products, compares offers, and can complete some or all of a purchase on their behalf. It covers a spectrum: from an assistant that narrows down options and hands the person a link, to a feature like Amazon's Buy for Me that places the order autonomously. The term became mainstream through 2025 and 2026 as OpenAI, Google, Amazon, and Microsoft each launched their own version of it.
A handful of technical protocols coordinate this transfer of trust and payment: the Agentic Commerce Protocol (ACP) co-developed by OpenAI and Stripe, Google's Universal Commerce Protocol (UCP), and the Agent Payments Protocol (AP2) that authorizes an agent to pay. On the purchase itself, Agentic Checkout and the payment networks' Agentic Tokens are what let a transaction close safely inside, or just after, a conversation with an agent.
In Europe, a Sopra Steria study estimated 310 billion EUR in AI-assisted ecommerce transactions over the next ten years, with 55% awareness of agentic commerce among consumers. In France, a Fevad x KPMG 2026 study found that 31% of online buyers already use generative AI to prepare a purchase. As of October 2026, the dominant pattern is still discover in the AI agent, buy on the merchant's own site, rather than a fully autonomous checkout: no audited top 100 French ecommerce site allows a fully autonomous payment yet, according to a September 2026 VISIPLUS academy study, and OpenAI itself pivoted away from in-chat purchase completion in March 2026.
For GEO, agentic commerce matters doubly: a brand first needs to be the product an agent discovers and recommends, which is where AI Shopping visibility work overlaps with Agentic Browsing and Computer Use, then needs its checkout and payment infrastructure ready for the agent to complete or hand off a purchase. See the Agentic Commerce in 2026 article and BotRank's AI Shopping feature.
Examples
A shopper on Amazon enables Buy for Me inside Alexa for Shopping: the assistant finds a product fitting the description, confirms price and delivery with the person, then completes the order autonomously inside the person's existing Amazon account.
A B2B buyer tells a procurement agent to replenish a shortlist of office supplies from pre-approved suppliers, below a set budget. The agent compares prices across the approved catalog and places the orders without a human clicking through each checkout.
Frequently Asked Questions
What is the difference between agentic commerce and simple AI-assisted shopping?
AI-assisted shopping is a human reading recommendations from an AI chat and then buying somewhere themselves. Agentic commerce goes further: the agent itself completes part or all of the transaction, using protocols like ACP, UCP, or AP2 and a payment method the person has authorized in advance.
Can AI agents already complete a full purchase autonomously in Europe?
Only in isolated cases. Worldline, ING, and Mastercard demonstrated a production transaction where an agent autonomously bought a concert ticket, but no audited top 100 French ecommerce site allowed a fully autonomous payment as of September 2026, and most activity still ends with the person completing checkout on the merchant's own site.
Why did OpenAI scale back Instant Checkout?
In March 2026, OpenAI said Instant Checkout is moving to Apps, pulling in-chat purchase completion back in favor of connected merchant apps. The stated reason: people search for products inside ChatGPT but rarely complete the purchase there, and the merchants that had enabled native payment were still few.
What should a brand do to prepare for agentic commerce?
Make sure product data is structured and accurate enough for an agent to compare, keep the brand visible and well cited where agents search, the core of GEO, and treat checkout and payment readiness, including support for protocols like ACP, UCP, or AP2 where relevant to the brand's market, as a separate, later-stage project.
