On 17 July 2026 France’s Autorité de la concurrence published Opinion 26-A-05 on the competitive functioning of the AI agents sector — the first opinion by a competition regulator to examine agentic commerce directly. It is a policy document, not a merchant guide, but three of its findings are operational for anyone selling online.

Under 5% of e-commerce traffic today, 20–25% by 2030

The Autorité’s own framing: traffic “currently redirected directly to e-commerce websites from AI agents remains minimal (less than 5%),” but “could become a significant channel for reaching consumers by 2030 (between 20% and 25%).”

Two things make that number worth more than the usual forecast. It comes from a regulator with subpoena-adjacent information-gathering powers, built on roughly 40 contributions to a public consultation. And it is a traffic projection, not a sales one — the share of visits arriving through an agent rather than a search box.

The market that generates it is concentrated: OpenAI, Google and Anthropic together hold “more than 84%” of it.

The barrier it names for merchants is missing standardised data

The regulator identifies what stands between agentic commerce and scale, and for the merchant side it is not consumer trust or payment rails. It is data:

Agentic commerce is also subject to a number of specific barriers, particularly technological barriers (such as the lack of standardised data or established protocols) and behavioural barriers.

That is a competition authority independently arriving at the thing that shows up in every store audit. A product page an agent cannot parse into an identifier, a price, a currency and an availability state is not a slow-converting page — it is an absent one.

The Autorité also flags a second-order cost merchants tend to miss: “when purchases are made through an AI agent, e-commerce websites lose access to users’ behavioural data, which could lead to a deterioration in the quality of the service they provide.” The sale can complete and still leave you blinder than before.

Who ranks you inside an agent is now a regulatory question

The sharpest part for merchants is what the Autorité puts under watch. Recommendation 2 calls for vigilance on “the definition and implementation of the key parameters that may influence how AI agents select, rank or recommend the services.” It names the risk plainly: discrimination and self-preferencing, “the conditions imposed on e-commerce websites for their content to be made available to users through an AI agent’s interface,” and the fact that “advertising and potential partnerships may also influence the responses provided to users.”

Recommendation 6 goes to the standards themselves — those governing agentic commerce “should be developed and maintained through transparent, open and collaborative processes,” with the highest openness compatible with security. The regulator’s concern is that new entrants must “comply with standards they did not help establish.”

For a merchant, that reads as an honest warning: the ranking layer is opaque, contested, and not yet regulated. Nobody is going to hand you a fair-visibility guarantee this year.

What that leaves a merchant able to control

If selection criteria are opaque and the standards are still being written, the only lever wholly on your side of the line is being unambiguously machine-readable and protocol-reachable — so that whatever an agent’s ranking parameters turn out to be, you are not disqualified before they apply.

That is the difference between the two failure modes we keep separating: an agent that ranks you below a competitor made a choice; an agent that cannot resolve your price, variant or checkout never had one to make. Only the second is entirely yours to fix. Our protocol-adoption data across 27 supplement stores shows how few stores have closed it, and the two checkout rails show where the standards work is actually happening.

The regulator’s timeline gives that work a deadline; being unreadable when 20–25% of your traffic arrives by agent is not a ranking problem, it is a disqualification.

Sources