Agentic Commerce in 2026: Where Do Europe and France Stand?

A shopper no longer necessarily types "best cordless vacuum" into Google. Instead, they ask ChatGPT, Gemini, or Microsoft Copilot to compare three models, check stock at a retailer and, increasingly, complete the purchase itself. This is agentic commerce: an AI agent that searches, compares, sometimes negotiates and buys on behalf of a human, with or without final confirmation. The topic became central in 2026, driven by OpenAI's much-discussed retreat from Instant Checkout, by Google's launch of the Universal Commerce Protocol and by a widely covered opinion from France's competition authority on the risks the sector poses.

This article takes stock of what agentic commerce actually is, the protocols and players shaping it, and its real level of adoption in Europe and France as of September 2026. The goal is to give brands a factual basis for deciding what to do next, without either overreacting or dismissing the shift.

What is agentic commerce?

Agentic commerce refers to transactions in which an AI agent acts on behalf of a user throughout the purchase journey: searching for products, comparing them, checking availability and price and, in some cases, paying, without the human visiting each merchant site directly. The agent can be a general-purpose conversational assistant (ChatGPT, Gemini, Copilot, Claude), an assistant built into a marketplace (Rufus, now folded into Amazon's Alexa for Shopping) or a specialized agent a company builds for its own procurement needs.

Three often-confused concepts should be kept separate:

• Conversational commerce: a chatbot that answers questions and points to a product page, without automating the transaction itself.
• AI-assisted shopping: an engine such as ChatGPT Shopping or Google AI Mode that recommends products within its answer, with a click through to the merchant site to complete the purchase.
• Agentic commerce proper: the agent holds an authorization, one-time or standing, to trigger payment itself on the user's behalf.

In 2026, most of the observed volume still falls into the first two categories. True agentic commerce, with fully autonomous end-to-end payment, remains marginal, as the French data detailed below confirms. On trust and autonomous payment specifically, we published a dedicated article: AI shopping, trust and autonomous checkout.

How does agentic commerce actually work?

For an agent to reliably find a product, verify its price and trigger a payment, agent, merchant and payment network need a shared language. Several open protocols emerged in 2025 and 2026, backed by competing companies that nonetheless converge on the same need for standardization.

Protocol	Backed by	Role
ACP (Agentic Commerce Protocol)	OpenAI and Stripe	Checkout standard used by ChatGPT to connect an agent to a merchant's payment system.
UCP (Universal Commerce Protocol)	Google, with Shopify, Etsy, Wayfair, Target and Walmart	Common language between agents and online stores to browse a catalog and trigger a purchase in AI Mode and the Gemini app, through Google Merchant Center.
AP2 (Agent Payments Protocol)	Started by Google, donated to the FIDO Alliance	Secures the payment itself, including "human not present" purchases triggered under a pre-authorization.
MCP (Model Context Protocol)	Anthropic	Generic connection layer between a model and external tools, including merchant systems.
TAP (Trusted Agent Protocol) and Agent Pay	Visa and Mastercard	Let card networks recognize an agent as a trusted buyer and secure the transaction on the card side.
TACP (Trusted Agentic Commerce Protocol)	Forter	Fraud detection tailored to transactions initiated by an agent rather than a human.

Two layers are worth distinguishing: the product and catalog language (ACP, UCP) and the payment layer (AP2, TAP, Agent Pay). A merchant does not need to pick a side: the major retailers named by Google under UCP also work with ACP, and Stripe processes transactions coming from both standards.

The players and their strategies in 2026

Every platform is moving at its own pace, with sometimes opposite choices on how much room to give autonomous payment.

OpenAI: from momentum to caution

OpenAI launched Instant Checkout on September 29, 2025 in the United States, before expanding it to Europe in late August 2026. In-chat one-click payment was nonetheless scaled back in the first quarter of 2026: Walmart measured a conversion rate roughly three times lower for a purchase completed inside the conversation compared with an outbound click to the merchant site, despite attracting roughly twice as many new customers as classic search. OpenAI has since refocused ChatGPT Shopping on product discovery; the details of product selection and the merchant feed are covered in our guide on how to get your products into ChatGPT Shopping.

Google: infrastructure rather than a chatbot

Google launched the Universal Commerce Protocol in January 2026, then expanded it in March with cart management and catalog access. The distinctive part of the approach is that it runs through Google Merchant Center, already used by millions of merchants for Shopping Ads, which becomes the control center for making a catalog eligible for purchase in AI Mode and the Gemini app through the native_commerce attribute. Building on existing infrastructure rather than an isolated chatbot is likely the most structurally significant strategy for merchants already present on Google Shopping.

Microsoft, Amazon and Perplexity: three different playbooks

Copilot offers an integrated checkout for part of its catalog, in a logic close to Bing Shopping. Amazon merged its Rufus assistant with Alexa+ in May 2026 to create Alexa for Shopping, betting on purchase history and proprietary marketplace data rather than an open protocol; CEO Andy Jassy justified the move by explaining that shoppers who want to buy from a specific retailer will start with that retailer's own assistant. Perplexity, for its part, relies on a free merchant program to attract catalogs rather than entering a power struggle with the major payment networks.

The 2026 turning point: "discover in AI, buy on site"

The Instant Checkout story has become the most cited illustration of a broader pattern. After months of testing, the industry is converging on a model where AI serves as a discovery and comparison engine, while the transaction itself is completed on the merchant's own site or app, an environment where the merchant keeps control of the experience, customer data and after-sales relationship. This model, summed up as "discover in AI, buy on site," explains why product data, technical specifications and presence in AI engine answers now matter more than integrating a one-click buy button.

For a brand, the practical consequence is direct: the battle is not yet about integrating a payment system into a chatbot, but about the visibility and accuracy of product information that agents read before recommending one brand over another. That is exactly the ground covered by BotRank's AI Shopping feature, which tracks the presence, win rate and average rank of products in AI engine carousels.

Where does Europe stand?

A Sopra Steria study conducted in February 2026 among 8,400 consumers across eight countries (France, Belgium, Germany, the United Kingdom, Italy, Spain, the Netherlands and Norway) offers a detailed picture of Europe's maturity on the topic.

Indicator	Result
Ten-year market potential	€310 billion in European e-commerce transactions could be assisted by AI agents
Awareness of agentic commerce	55% have heard of it, only 13% describe themselves as truly familiar with it
Trust in current providers	41% trust no existing provider to handle a delegated purchase
Most legitimate provider	27% name banks first, ahead of technology platforms
Categories delegated most readily	45% for electronics, versus only 16% for health and groceries

Awareness of the topic varies sharply by country: it reaches 76% in Norway and 68% in the Netherlands, against only 38% in France. The gap between the interest expressed in agentic commerce's potential and the actual trust placed in the providers offering it is the most useful data point for a brand: it shows the barrier is not technological but relational.

On the regulatory side, France's competition authority (Autorité de la concurrence) published an opinion (26-A-05) on July 17, 2026 that became a reference point across the European Union. It noted that a small number of players control most of the AI agent sector and flagged three risks: self-preferencing in the recommendations of an agent built into a search engine or marketplace, opacity in the criteria used to rank the products an agent presents, and the loss of the customer relationship for merchants whose sale is completed entirely inside a third party's interface. The authority estimates that AI agents currently direct less than 5% of traffic to merchant sites, a figure it projects will reach 20% to 25% by 2030, a level comparable to today's SEO or paid search.

Where does France stand?

Data published in September 2026 shows a French market that remains cautious, yet is already engaged on the consumer side.

Indicator	Result	Source
Online shoppers already using generative AI to prepare a purchase	31% of online shoppers, up to 73% among regular generative AI users	Fevad x KPMG, 2026
Top 100 French e-commerce sites with an AI assistant deployed	26% (22% across a broader panel of audited sites)	VISIPLUS academy, Sept. 2026
Sites letting an agent add items to a cart automatically	A little over a third	VISIPLUS academy, Sept. 2026
Sites letting an agent complete payment fully autonomously	None of the audited sites, to date	VISIPLUS academy, Sept. 2026
Trust in an AI assistant for pre-purchase advice	47%, versus only 30% at the payment stage	VISIPLUS academy, Sept. 2026
Preference for human contact during the purchase	66% of buyers	VISIPLUS academy, Sept. 2026

France's e-commerce market was worth €196.4 billion in 2025, up 7%, across 3.2 billion transactions. It is on this already solid base that the agentic question builds: according to Fevad and KPMG, the goal is no longer simply to earn a click but to be retained in a recommendation an AI formulates, which shifts part of the value toward the players that control the protocols, the API connections and the structured product catalogs.

This overall picture matches what we see in our own product data: AI-assisted discovery is moving fast on the consumer side, while transactional autonomy is moving far more slowly on the merchant side. That gap is a window of opportunity rather than a reason to wait: brands that structure their product data now gain an edge in a recommendation mechanism that is being built today, well before autonomous payment becomes mainstream.

What agentic commerce changes for brands

Three concrete shifts follow from the above.

Product data becomes the new competitive battleground

An agent does not browse a product page the way a human does: it reads a structured feed (OpenAI's Product Feed, Merchant Center catalog, Shopify data) and schema.org markup. An inconsistent price between the feed and the page, a missing image or a missing product identifier is enough to exclude an item from recommendations, even when its classic SEO is excellent.

Measuring presence becomes a new discipline

Knowing that a product ranks well on Google says nothing about its presence in ChatGPT Shopping, Gemini or Perplexity carousels. That is exactly what BotRank's AI Shopping feature measures: presence rate, win rate against competitors, average rank and identification of the merchants actually cited, including when they are resellers rather than the brand's own site.

The customer relationship is decided at the point of sale

As long as payment is completed on the merchant's own site, the merchant keeps the browsing data, the customer history and control of after-sales service, a point France's competition authority explicitly flagged as a competitive concern. That is one more argument for investing in upstream visibility rather than rushing to integrate a buy button inside a third party's interface.

How to prepare for agentic commerce: the checklist

• Audit how readable the product catalog is for an agent: an up-to-date feed, with consistent price and stock across the feed, the product page and structured data.
• Check that robots.txt does not block AI engine crawlers (OAI-SearchBot, Google-Extended, PerplexityBot) from product pages.
• Document customer reviews and third-party sources that mention the product; they carry as much weight as the merchant feed in an agent's selection.
• Measure current presence in the answers of the main AI engines before investing in a checkout protocol.
• Track announcements from OpenAI, Google and the payment networks on ACP, UCP and AP2, without rushing into a technical integration while real consumer usage remains limited.
• Strengthen the direct customer relationship and proprietary channels, so as not to depend entirely on a third-party platform once agentic payment becomes mainstream.

BotRank's take

Agentic commerce is not yet at the stage where a brand needs to pick a payment protocol. It is at the stage where a brand needs to know whether it exists in the answers of the agents currently driving product discovery, before those same agents gain more transactional autonomy. This shift resembles what happened with organic search fifteen years ago, on a shorter timeline: the competition authority points to 2030 for a structural tipping point. We track this topic daily with our clients through the AI Shopping feature, and we will keep documenting every significant shift in protocols and usage here, across France and Europe.

FAQ

Is agentic commerce already a reality in France?

Partly. French consumers are increasingly using generative AI to prepare a purchase, but no merchant site audited by VISIPLUS academy in September 2026 let an agent complete a payment fully autonomously. The current phase is assisted discovery and recommendation, not yet delegated payment.

What is the difference between ACP and UCP?

ACP (Agentic Commerce Protocol) is backed by OpenAI and Stripe and mainly serves ChatGPT. UCP (Universal Commerce Protocol) is backed by Google and relies on Google Merchant Center to connect merchant catalogs to AI Mode and the Gemini app. Both aim at the same goal through different ecosystems.

Why did OpenAI scale back Instant Checkout?

Early conversion data, notably at Walmart, showed that a purchase completed inside the conversation converted markedly worse than a click through to the merchant site, while attracting more new customers. OpenAI has since refocused ChatGPT on product discovery rather than one-click payment.

Does agentic commerce apply to every product category?

No. European studies show consumers delegate electronics and technical products more readily than health or grocery items, categories that require a higher level of trust.

How can a brand measure its presence in agentic commerce?

By tracking its presence, win rate and average rank in AI engine product carousels, as well as which merchants are actually cited for its listings. That is what BotRank's AI Shopping feature measures.

Sources

• Sopra Steria, study on trust in agentic commerce across Europe, May 2026
• Autorité de la concurrence, opinion 26-A-05, July 17, 2026
• E-commerce Nation, analysis of the competition authority's opinion
• Fevad x KPMG, agentic commerce study, 2026
• Ecommercemag, summary of the KPMG x Fevad study
• VISIPLUS academy, where agentic commerce stands in France, September 2026
• Google Cloud, introducing the AP2 protocol
• Google, donating AP2 to the FIDO Alliance
• Google Developers Blog, how the Universal Commerce Protocol works
• GeekWire, Amazon merging Rufus and Alexa+
• Digital Applied, analysis of the "discover in AI, buy on site" shift
• Stripe and OpenAI, announcing the Agentic Commerce Protocol and Instant Checkout