As discovery shifts from traditional search towards AI platforms, what does the move from SEO to AI-optimised visibility actually involve, and where should teams start?

It changes more of the mechanics than most people expect. Classic SEO gives you a keyword with a known search volume and a page to optimise against it, whereas LLMs withhold all of that — the keywords, the volumes, the search behaviour. Even the input looks different, since a ChatGPT prompt runs to about 23 words, where a Google query gets three or four, and the output narrows dramatically as you move along the shelf: 200-plus SKUs in store, three on a mobile screen, roughly eight in an AI answer. Once the shelf is that short, the only question left is whether you are in the answer at all. We break the work into three jobs that build on each other. Being found means AI systems can read and understand our products in the first place. Being trusted means the sources AI leans on are credible and mention us. Being chosen means there is an explicit reason to recommend us over the product next to us. The starting point has to be measurement, because the platforms hand you none of it. We built our own prompt set from autocomplete data, ran it repeatedly, and tracked which sources the answers cited — and the finding was uncomfortable. Retailers and comparison platforms consistently ranked above our own corporate and brand sites. AI trusts the retailer, which told us where to spend our energy: win the retailer shelf first, because that is the shelf the machine reads. The one caution I would add is to keep visibility as a diagnostic and sales as the KPI, because it is very easy to build a dashboard that climbs steadily while the P&L stays exactly where it was.

What does it take to design interfaces that AI agents can interpret and act on for comparison and checkout — and how far are most brands from being ready?

The technical requirements are surprisingly mundane. Agents want structured, crawlable, continuously updated product data that lends itself to comparison, which happens to describe precisely why retailer pages are so much easier for AI to read than brand pages are. We learned this the direct way on our own site: structured product information gets read, indexed and recommended, while the same content tucked into an accordion menu stays invisible. AI usability and user usability turn out to be two separate disciplines. That tension is where the real decision sits, and we made ours deliberately. We separate AI traffic from human traffic in our reporting, and in our German shop human traffic still clearly dominates, so we optimise the UX for humans and implement the AI improvements that leave that experience intact. Rebuilding a store around a traffic source that is still in the minority strikes me as premature, and I would be sceptical of anyone claiming they have already done it. As for readiness, most brands are further away than they think, and in my experience the gap comes down to ownership rather than technology. We set up a cross-functional taskforce spanning eCommerce, marketing, PR, media, IT and legal, precisely because the answer to "how does AI describe our brand" lives in six departments at once, which in practice means it lives nowhere.

If a brand becomes largely discovered and transacted through AI intermediaries, what happens to brand differentiation, and how do you protect it?

Our brands are the most valuable thing we have, so this is the question I spend the most time on, and I would say openly that it remains unsolved — ours included. Over the past decades brand manufacturers became genuinely good at the funnel. Every stage had its own asset in the right format in the right place, from awareness through consideration and intent to loyalty. That entire architecture now collapses into a prompt. The journey compresses inwards towards price, specification and sentiment, and what finally reaches the shopper is a few lines of text arriving at the exact moment the decision gets made. Which leads to the honest challenge: how does a brand owner transmit emotion and a brand core in three lines of machine-generated answer, text only? Everything we built our emotional differentiation on gets stripped out at the point of choice. What stays within our control is what the machine reads. The reasons to choose us have to be explicit and comparable — occasion, benefit, ingredient, use case — written down in a form AI can quote back. Third-party signals carry more weight than owned ones, so reviews, publishers, creators and retailer content matter more than they used to. And consistency becomes commercially relevant, because describing a brand the same way across five retailers gives AI one stable story to repeat. Get this wrong and the outcome is easy to picture: brands turn into specifications, and specifications compete on price.

What are you most looking forward to about joining eTail Germany Connect with other senior eCommerce, digital and marketing leaders?

It has always been an event that connects the two things that matter to me, learning and exchange, and the format leaves genuine room for both. Because the sessions run under the Chatham House Rule, people talk about what actually happened, including the parts that went wrong, which is exactly why I am glad to take part again.