Footprint, not fame: why AI ignores famous brands, and what everyone's still missing

We ran 14 UK fashion brands through the AI shopping engines quietly rewriting product discovery — ChatGPT and Perplexity. We asked the questions a real shopper types. The famous names were invisible. Here's what actually predicted visibility, measured on our own data.

Being famous doesn't make you visible

The first finding is the one that surprises people. Brand fame — the thing every marketing budget points at — does almost nothing for AI discovery.

Victoria Beckham has 33 million Instagram followers. Stella McCartney has 7 million. Both are household names with decades of press behind them. On non-branded shopper queries — where someone describes what they want rather than naming a brand — both scored zero. The AI never surfaced them.

Take the most-followed brand in the study. Victoria Beckham has 33 million Instagram followers — nearly five times the next-biggest following here — and one of the most recognised names in fashion. This is what our scan finds when it reads her product pages.

Second-lowest AI-readiness score of the fourteen, and no discovery visibility at all. Her product pages carry almost none of the structured data an agent needs to read and recommend them: no product schema, no semantic tags, barely any alt text. Fame does not translate into AI visibility. The data work is simply undone — and it is undone right across the category, which is precisely why this is an opening rather than a verdict. Whether closing it moves discovery is the thing we are now measuring.

Meanwhile the brands that broke through were smaller, less famous, digitally native: Lucy & Yak, Nobody's Child. Not the names you'd predict. Followers didn't predict visibility. Domain authority didn't predict it. When we tested each signal against AI discovery across all 14 brands, three of the four came back flat — no relationship at all.

One thing predicted visibility: web footprint

The signal that mattered was traffic — the size of a brand's web footprint. Brands with real footprint surfaced; brands without it didn't. The relationship was strong and statistically significant, and held whichever way we measured it.

The gap is stark. Brands with any discovery visibility had a median of roughly 1.5 million monthly visits. Brands with none sat around 280,000 — about five times smaller.

This makes sense once you understand how these engines work. ChatGPT and Perplexity don't recall brands from memory. They run a live search, read what comes back, and answer from it. So visibility isn't about who's biggest or best known — it's about whether you're in the documents the AI retrieves. Footprint is what puts you there.

The results in full

Product discoverability in AI = how often the brand surfaces on non-branded shopper queries. AI-readiness = overall Catalign scan score, 0–100, a weighted composite of seven catalogue dimensions. Domain authority = backlink-strength rating, 0–100. Web traffic = estimated monthly visits. AI-readiness figures updated September 2026, when feed fields moved from unscored to measured.

Read down the discovery column and the pattern is plain: the brands that surface sit at the top on traffic; the famous-but-low-traffic names sit at zero. And not one brand has strong data — the best manages the mid-forties out of a hundred.

The category is barely discovered at all

Here's the part that should get every ecommerce lead's attention. We split each brand's score in two — visibility when a shopper names the brand, versus visibility when they describe what they want.

Every brand scored well on its own name. Ask the AI about RIXO and it describes RIXO. That's table stakes. But on the queries that drive new customers — the category questions, the "best X for Y" searches — seven of the 14 brands scored exactly zero. The rest managed single digits. Ask the question a shopper would really type, and half these brands do not exist to the AI.

The whole category is undiscovered. That's not a failure. It's an opening.

Marks & Spencer, and the engine nobody is running

Marks & Spencer was the exception. It dominated discovery — appearing roughly four times more often than any other brand on non-branded queries.

Now look at its data. M&S scored 44 out of 100 overall. A D−. Its product pages carry clean schema markup and usable alt text, but no semantic tags for fit or fabric or occasion, no application templates, and only half the crawler configuration a shopping engine looks for. On the measures that let an AI understand what a product is and when to recommend it, M&S is as thin as everyone else.

So M&S isn't winning on good data. It's winning on brute footprint — 40 million monthly visitors, decades of editorial presence. But look closer at what it does get right: product schema at 90 out of 100, clean and complete, and the strongest alt text in the study at 62, even as the semantic layer sits at zero. And when we read the actual AI responses, M&S's own product pages were cited directly, again and again. That's the tell — strong structured data is what let its pages be retrieved and recommended. M&S is accidentally running half the second engine, and it's already paying off.

Which raises the obvious question. What happens to a brand that pairs footprint with data built properly for AI retrieval? Nobody in this category has done it.

The numbers are starker than "nobody has done it well". Semantic tags — the fit, fabric and occasion signals an agent needs to match a product to a query — averaged 0.7 out of 100 across the cohort, and only four of the 14 brands scored above zero at all. Alt text averaged 19 out of 100. Not one brand in the category reached 47 out of 100 overall.

What these brands do have is schema: 67 out of 100 on average. The foundations are in place. The layer that tells an agent what a product actually is — and when to recommend it — is not. That lever is sitting untouched across the entire mid-market.

Two engines, and only one is running

AI discovery runs on two things working together.

Footprint gets you into the conversation — into the set of documents the AI retrieves when it searches. It's built through coverage, content and time, and it's largely outside a single quarter's control.

Data structure is what lets the AI understand and recommend you once you're in front of it — the difference between a page it can parse into "machine-washable cashmere jumper, £120, UK-made" and a page it skips because it can't tell what it's looking at.

Right now the category runs on the first engine alone, and unevenly. M&S has footprint and mostly-poor data. Everyone else has poor data and not enough footprint. Nobody is running both. The brands that fix that first — that get their content into the sources AI cites and structure it so the AI can retrieve it — will own AI discovery in their category before their competitors register it's a channel.

What Catalign does

We measure your real AI discovery visibility across ChatGPT and Perplexity — the number most brands badly overestimate, because they assume being known means being found. It doesn't.

We map exactly which sources the AI cites in your category, and where you're absent — so you know precisely where your content needs to show up.

And we enrich your product data so that when the AI looks, it can find you, understand you, and render you as a rich card with imagery, price and ratings — not a line buried in a paragraph.

Footprint is the long game. But the data structure that makes your own pages retrievable — the way M&S's schema made its pages citable — is the part you can fix now. It's the fastest lever available, and right now nobody in the category is pulling it.

About this study: we measured 14 UK mid-market fashion brands across ChatGPT and Perplexity in August 2026, using a fixed set of category and branded shopper queries per brand. Findings are directional — a sample of this size shows patterns, not certainties — and we're continuing to test whether enriching a brand's data measurably moves its discovery visibility.