How Premium Brands Get Found in AI Search with Christine Ullmann of John Lewis

Search is changing what discovery means for retail brands.

That is true for every retailer, but it is especially true for premium retailers. If you are not trying to win every shopper on price, visibility alone is not enough. The question becomes: visible to whom, in what context, and with what story attached?

That was the focus of my conversation with Christine Ullmann, Head of SEO & Organic Growth at John Lewis.

For those outside the UK, John Lewis is one of those brands that needs a little context. It is an iconic British retailer with a long history, a premium positioning, and a reputation built around quality, service, trust, and customer care. That makes it a fascinating case study for where search is going, because AI-assisted discovery is not only about matching a page to a keyword. It is increasingly about understanding what a brand stands for, who it is right for, and why someone might choose it.

Christine has spent more than 15 years working across ecommerce, travel, and retail brands, and she also has a PhD in Social Anthropology. That background came through throughout the conversation. She does not talk about search as a spreadsheet of keywords. She talks about search as a way to understand people, culture, demand, trust, and intent.

That felt like the right lens for this topic.

Key takeaways

• Visibility is too vague on its own. Premium retailers need to ask what they are visible for, who they are visible to, and what brand meaning comes with that visibility.
• Keywords still matter, but they are no longer enough. AI-assisted search makes every query more contextual, more personal, and harder to reduce to search volume alone.
• Brand, entities, sentiment, and third-party mentions now shape how discovery systems understand you. What others say about you matters, not only what you publish on your own site.
• Product detail pages are becoming more important as discovery surfaces, especially through product search, Merchant Center, AI answers, and agentic shopping.
• Retailers need to test metrics, not only tactics. Rankings, traffic, visibility, and impressions still exist, but they do not always mean what they used to mean.
• Testing helps teams avoid harm as well as find growth. A negative result can save the business from rolling out something damaging.
• SEO becomes more powerful internally when it is framed around customers, commercial value, and uncertainty, rather than channel jargon.

Optimize for your audience, not just your keywords

Christine started with a point that sounds simple, but becomes more important the more search changes: you need to understand who you are, who your target audience is, and what you are to them.

That is especially important for a premium retailer.

A search like "black dress" could mean almost anything. It could be someone looking for the cheapest possible dress for one event. It could be someone looking for something better made, longer lasting, more sustainable, or suitable for a particular occasion. Those are not the same shopper, even if the keyword is identical.

Christine's point was that a premium brand should not be afraid to signal clearly who it is for. If someone wants a one-night purchase that they will throw away or donate immediately afterwards, John Lewis may not be the right answer. If someone wants quality, service, sustainability, and trust, then John Lewis needs to make those signals unmistakable.

That is a useful shift in framing. Traditional SEO can make us think in terms of ranking for the keyword. AI-assisted discovery pushes us to think in terms of being the right answer for the right person.

In Christine's words, the work becomes less about optimizing for keywords and more about optimizing for your audience.

Search volume is becoming less useful than it used to be

I have been using the phrase "search volume one" recently, and this conversation was a good example of why.

Even when someone types a head term like "black dress", the system increasingly has more context than that phrase alone. It might understand where the person lives, what they have searched before, what they tend to buy, what brands they trust, what stores are nearby, and what kind of trade-offs they usually make.

Once that context gets folded into the result, the "same" query is not really the same query anymore. The words might be the same, but the intent is personalized.

That does not mean keyword research is dead. It does mean search volume is no longer the clean proxy for demand that we sometimes pretend it is.

Christine put it well: the metrics we know still exist, but they do not mean the same thing anymore. Search volume, visibility, rankings, and traffic all still have a role, but we need to ask whether they are really measuring the value we think they are measuring.

That is a big mental shift. We should not only experiment on tactics. We should experiment on metrics too.

The wider digital ecosystem now matters more

One of the strongest parts of the conversation was Christine's explanation of context and sentiment.

We have all spent years thinking about links. Links still matter. PageRank is not dead. But the old way of thinking about link metrics, anchor text mixes, C-blocks, and directory links is not the centre of the game anymore.

What matters more now is the meaning around the mention.

If John Lewis is mentioned regularly by high-end publishers, in articles read by a certain audience, with language around quality, service, sustainability, or trust, that creates a different kind of signal. It helps search engines and LLMs understand not only that John Lewis exists, but what John Lewis is associated with.

Christine gave a practical example: if a publisher says that John Lewis will deliver your oven and recycle your old one, that is not just content. It is a relationship between entities and attributes. It connects John Lewis with ovens, delivery, recycling, service, and added value.

That matters because AI systems and search engines are increasingly trying to interpret relationships, not only match words. They are looking at entities, sentence structure, brand mentions, publisher context, and audience fit.

For premium brands, that means the story around the brand has to be consistent across the whole digital ecosystem. Your website matters. So do publishers, reviews, product data, social, local listings, and the way your products appear across shopping surfaces.

Omnichannel is becoming a trust signal

For John Lewis, the physical store network is not separate from search. It is part of the value proposition.

Christine made the point that stores used to sit in the local SEO bucket. You optimized Google Maps. You checked listings. You made sure the basic local data was correct.

In an AI-assisted world, that physical presence becomes much richer.

If someone's oven breaks and they ask what to do, the answer may not be a blue link. It might be a recommendation that takes into account their location, whether they drive, whether a store nearby has the product on display, whether there is parking, whether installation is available, whether the retailer offers an extended warranty, and whether they will recycle the old appliance.

That is a very different search experience.

It also shows why premium retailers need to think beyond pages. If your service, stores, warranties, returns, advisers, cafes, delivery options, and aftercare are part of why customers choose you, those signals need to be available to the systems that are shaping discovery.

That is not only an SEO problem. It is a product data problem, a local data problem, a brand problem, and a customer experience problem.

Product pages are becoming more important

We also talked about the shift from category pages to product detail pages.

For years, ecommerce SEO work often revolved around PLPs, category pages, and listing pages. Those still matter. But product search, Merchant Center, AI answers, and shopping experiences are pushing more discovery directly toward PDPs.

That changes what a good product page needs to do.

A PDP is not only a conversion endpoint anymore. It may be the first meaningful experience a customer has with your brand. It may also be the page an AI system uses to understand whether a product is right for a particular customer.

This is where Christine made a useful point about not putting everything on the same page.

For mobile users especially, dumping every possible detail into the product description can be overwhelming. In many cases, it is better to show the key decision-making details on the page, then link to supporting fact sheets, guides, or resources.

For example:

• A beauty product might need clear signals around whether it is vegan, cruelty-free, suitable for sensitive skin, or fragrance-free.
• An appliance might need energy efficiency, dimensions, installation information, delivery options, and recycling details.
• A fashion product might need sizing, materials, fit, care instructions, styling inspiration, and reassurance that it will look good in the situations customers care about.

The bigger point is that product pages should not be treated as isolated pages. They should be connected into a broader information system.

Move from strict taxonomy to hub-and-spoke thinking

Christine also challenged the classic ecommerce pyramid.

Most ecommerce sites are built around a taxonomy: department, category, subcategory, product. That structure still has value. Nobody is saying retailers should rip it up.

But customer journeys do not always follow that neat pyramid.

Someone decorating a living room might need products from several different categories: sofa, cushions, lighting, rugs, storage, wall art, paint ideas, and styling inspiration. Those items may sit in different parts of the taxonomy, but from the customer's point of view, they belong together.

That is where hub-and-spoke thinking becomes useful.

You can keep the ecommerce taxonomy while also building thematic hubs that connect adjacent needs. These might be based on rooms, occasions, styles, life moments, values, or practical customer problems.

The key question is: how do we help the customer, or the customer's AI assistant, make the connection?

Sometimes the answer is information. Sometimes it is inspiration. Sometimes it is a product. Often it is all three.

That is why the old funnel model has always been a little too tidy. People do not move cleanly from awareness to consideration to purchase. They browse, compare, get inspired, leave, return, ask friends, watch videos, walk past shop windows, and sometimes buy something completely different from what started the journey.

Digital discovery is catching up with that messy reality.

Social, visual discovery, and AI search are converging

One of the most interesting parts of Christine's view is that she does not isolate SEO from the rest of discovery.

She talked about TikTok search, visual search, Circle to Search, Pinterest-style inspiration, and the role of social content in brand discovery. Her point was not that every social video needs to be "SEO optimized" in some awkward way. It was that customers are already searching in many places.

A person looking for living room ideas may not start with Google. They may start with TikTok, Pinterest, Instagram, YouTube, or an AI assistant. They may not even think of what they are doing as search. But they are still discovering.

For a retailer like John Lewis, this creates an opportunity. The imagery from a Christmas campaign, a room set, or an editorial feature can do more than create a brand impression. It can become part of how customers discover products, styles, and ideas.

The commercial impact may not be immediate. Someone might not buy a cushion on the first visit. But they may remember where the inspiration came from. They may come back later when they are ready to buy.

That matters because brand and search now flow in both directions. Search visibility creates brand impressions, and brand strength changes what people click when they see you.

Testing should help us make better decisions, not only give yes-or-no answers

One of the themes I brought into the conversation was multi-metric testing.

In the simplest version of SEO testing, the system gives a thumbs up or thumbs down. More organic traffic is good. Less organic traffic is bad.

But that is not how many real decisions work now.

We are increasingly seeing tests where one metric moves positively and another moves negatively. A change might improve visibility but reduce click-through rate. It might help LLM traffic but hurt classic organic traffic. It might improve clicks but lower conversion rate. It might improve product search visibility but create a weaker user experience.

That does not mean the test failed. It means the result needs interpretation.

The value of testing is not only that it tells you what to do. It gives you a more credible basis for making a decision.

That distinction matters. The output is not always "roll this out" or "do not roll this out." Sometimes the output is: here is the trade-off, here is what we know, here is what we do not know yet, and here is the next test we should run.

Negative tests can be valuable because they avoid harm

Christine made a point that I think more SEO teams should take seriously: avoiding harm has value.

We often celebrate positive tests. We share the winners. We talk about uplift. Then when a test is negative, we file it away and move on.

But if a test prevents a damaging rollout, that is a business win.

If a change looked like a good idea, had stakeholder support, and would otherwise have gone live across thousands or millions of pages, a negative test may have saved a huge amount of traffic and revenue. That avoided loss should be part of the story.

This is especially important in ecommerce, where "best practice" can be dangerous. Something that worked in another industry five years ago may not work on your site, with your products, your customers, and your brand position.

Christine said she encourages teams at John Lewis to challenge her in the same way. Even if she says something based on years of SEO experience, it should still be treated as a hypothesis.

That is a healthy culture. It moves SEO away from authority and toward evidence.

Default to deploy: when neutral is good enough

There is another side to this as well.

Sometimes a business wants to make a change for non-SEO reasons. It may be better for users. It may be better for brand. It may support sustainability, accessibility, service, or trust. In those cases, the test is not asking: will this increase SEO traffic?

It is asking: will this harm SEO traffic?

At SearchPilot, we often call this "default to deploy". If the change is neutral or inconclusive, we ship it because the business wants it for other reasons. We only stop if the test shows meaningful harm.

That is a powerful frame for premium brands.

A retailer may want to highlight service and quality more than price. In the short term, that could reduce conversion rate among price-led shoppers. But if the goal is to attract and retain value-led customers, that may be the right decision.

Testing does not remove the need for judgement. It makes the judgement better informed.

Testing lets teams be bolder

A mature testing program changes the kinds of ideas people are willing to try.

Without testing, teams tend to propose ideas they feel confident will work. That usually means safer, more incremental ideas. With testing, you can try higher-variance ideas because you have a safety net.

That is how you get beyond checklist SEO.

Instead of only asking whether a page has two H1s or missing meta descriptions, you can ask more interesting questions:

What happens if we change the role of product descriptions?

What happens if we link to fact sheets instead of showing every detail on the PDP?

What happens if we add inspirational content to a PLP?

What happens if we treat a category page as a gateway to a topic, not just a grid of products?

What happens if we highlight service and warranty over price?

Those are not easy audit-tool recommendations. They are strategic questions. And they are exactly the kinds of questions that matter more in AI-assisted discovery.

Move from "what" to "why"

Christine made another point that is central to how I think about testing: the result of a test should push you from "what happened?" to "why did it happen?"

Say you read a SearchPilot case study where removing meta descriptions was positive. If you test the same thing and it is negative, that does not mean one result is "right" and the other is "wrong". It means the context differs.

The useful question is why.

Why did it work on one site and not another? Was it the page type? The search result presentation? The brand? The existing meta descriptions? The query mix? The intent? The role of snippets? The product set?

That is where testing becomes strategic. It helps you understand your own system rather than copy someone else's tactic.

The same applies to structured data. Adding schema to a PDP may support rich results, Merchant Center visibility, price visibility, ratings, and possibly AI discovery. Adding the same kind of structured data to a PLP may have much lower impact. The tactic is the same. The context is different.

SEO is becoming a people discipline

Christine said something that I strongly agree with: the more senior you get in SEO, the less it is about optimization and the more it becomes a people discipline.

That does not mean technical knowledge stops mattering. It means technical knowledge is not enough.

In a large retail organization, SEO touches engineering, product, brand, merchandising, experimentation, finance, data science, stores, and customer experience. If SEO is treated as a team that turns up with demands, it will struggle. If SEO becomes a shared way of thinking about customers, discovery, and business value, it becomes much more powerful.

Christine talked about developers who start out skeptical of SEO, then become advocates once they see how small technical changes can create measurable commercial impact. That is a familiar pattern. When an engineer can say, "I helped bring more customers into the business," the work feels different.

The same applies to brand teams. SEO and brand have sometimes been in tension, but they should not be. Brand work creates demand and trust. Search captures and reflects that demand. In AI search, brand signals may also shape whether you appear in the right context.

The more these teams understand each other, the better the work gets.

Speak to finance in finance language

Toward the end of the conversation, we talked about how to explain SEO value internally.

Christine's advice was direct: define value, and understand how the people you work with define value.

If you tell finance that clicks increased by 8%, the SEO team may be excited. Finance will ask how much money that made.

That does not mean you need perfect certainty before you speak to them. In fact, being clear about uncertainty often makes you more credible.

You might say: we have driven more clicks. Our early guardrail metrics suggest conversion rate has held up. Based on that, we can model a possible revenue impact. There are still assumptions, and here is what we are checking next.

That kind of framing works because finance teams deal with uncertainty all the time. Forecasts, weather, stock, demand, margin, seasonality, and trading conditions all involve assumptions. They do not need false confidence. They need clear thinking.

The same principle applies to brand teams, product teams, and leadership. Do not bring them SEO jargon and expect them to care. Bring them a story that connects to the value they already care about. If you've seen traffic fall while revenue held up, we've written about this directly.

Mixed teams make the analysis better

One of the final ideas Christine shared was about building respectful working relationships with people who think differently.

She described herself as someone with an anthropology background who understands audiences and people, but not as a pure statistics person. She works closely with experimentation colleagues who think differently, and that difference makes the work better.

I have seen this firsthand too. The best test ideas often come from combining perspectives. SEO sees one opportunity. Product sees another. Experimentation sees a measurement issue. Finance sees a modelling question. Brand sees a positioning risk.

When those teams trust each other, one result can turn into five better follow-up ideas.

That is the compounding effect of a strong experimentation culture.

What I want premium retailers to take away

If you are a premium retailer, the AI search challenge is not simply to show up more.

It is to show up in the right moments, for the right customers, with the right value signals attached.

That means being clear about who you are. It means making service, quality, sustainability, stores, advice, warranty, and trust visible across your digital ecosystem. It means thinking beyond keywords and beyond your own website. It means treating PDPs, product data, internal links, hubs, editorial, social, and third-party mentions as part of the same discovery system.

It also means testing more than tactics. Test whether your metrics still mean what you think they mean. Test whether your page changes create trade-offs. Test whether a change avoids harm. Test whether the things that matter to your brand are being understood by search systems and by customers.

AI search is not making SEO less strategic. It is making the strategic parts harder to ignore.

Put Search in Control Mode with SearchPilot

That is why controlled experimentation matters.

Search is often one of the biggest channels in the business, but it is also one of the least understood. In AI-assisted search, that uncertainty grows because discovery is spread across classic results, shopping surfaces, product feeds, LLM answers, social search, and personalized journeys.

SearchPilot helps enterprise retailers turn those uncertainties into measurable experiments. We test changes across category pages, product detail pages, navigation, content, and key commercial templates, then help teams understand what moved, what did not, and what the trade-offs mean.

For premium retailers, that matters because the right answer is not always "more traffic at any cost". Sometimes the right answer is better-qualified visibility. Sometimes it is protecting brand value. Sometimes it is proving that a values-led change does no harm. Sometimes it is avoiding a rollout that would have quietly damaged performance.

The future of search is not something you can predict from a conference slide or a best-practice checklist. You need to test it on your site, with your customers, against the metrics your business cares about.

Stop trying to predict how AI search will behave. Experiment to discover it. If you want tailored test ideas for your top PLPs and PDPs, schedule a demo and we will share a starter list and a clear path from validation to velocity to control.