A smoke test on the AI discoverability scoring engine we built at Geoffy returned something that looked like a bug: nine of ten brands scanned that morning scored zero on Perplexity. A zero-rate that high demanded a sanity check against the live product.
So we picked three of the brands — one footwear, one apparel, one furniture; all real businesses doing seven and eight figures, none fading, none obscure — and queried Perplexity directly with the obvious category-discovery question. No brand name in the query. Just the category and the buyer intent, the same prompt structure our engine sends.
Premium men’s footwear. Perplexity returned a confident top ten: heritage shoemakers, luxury houses, cult favourites. The test brand — a premium casual name most shoppers in its home market can name without thinking — did not appear.
Premium men’s lifestyle apparel. Ten brands surfaced, spanning coastal heritage labels and design-led newcomers. The test brand — serious revenue, a flagship store on one of the world’s most famous shopping streets, listicle-tier recognition for exactly this aesthetic — did not appear.
Premium direct-to-consumer bed frames. Ten names returned. The test brand — arguably the most visible DTC brand in its category over the last five years, the kind of design recognition that ends up in coffee-table magazines — did not appear.
Three categories. Three brands you have almost certainly seen advertised. Three zeros. And when we checked the engine’s scores against the live Perplexity product, they agreed: the brands really aren’t surfaced for these queries. The miss is structural, not a measurement error.
Why this happens
AI products don’t have an index the way Google has an index. When you ask Perplexity for “the best premium men’s footwear brands in 2026,” it isn’t ranking pages by keyword relevance. It reads a handful of recent editorial sources, synthesises what they collectively say, and serves you the names that appear most often across that small retrieval set.
If your brand isn’t named in those sources, you don’t appear. It doesn’t matter how recognisable you are. It doesn’t matter what you spend on Meta. It doesn’t matter that the question is one your buyers ask every day. The retrieval system doesn’t see you.
The brands that lose out aren’t the obscure ones. They’re the brands whose recognition lives in cultural awareness and paid social rather than in the editorial and structured corpus AI models actually draw on. Years of consumer-awareness building — none of it encoded where the machine looks.
And yet the buyer asking that question of ChatGPT or Perplexity or Gemini is exactly the buyer who would have considered these brands if shown the option. Clear product story, real customers, distinctive position — none of it helps if the retrieval doesn’t surface it.
The lesson
Visibility in AI commerce is not something you inherit from existing brand strength. It is its own discipline, with its own mechanics — being present, consistently and legibly, at every depth of buyer intent, from the broad question (“I need new slip-ons”) through the mid-specific (“waterproof leather slip-ons for travel”) to the precise (“premium slip-ons, vegan-leather upper, arch support”). That practice — making a brand the reference answer across those depths — is what we call Reference Engineering, and it’s the work Geoffy systematises.
If you run a premium brand, run the test yourself. Open Perplexity. Type the question your best customer would ask, without your brand name. See whether you appear. If you don’t, you have company — and a problem that won’t get smaller as a larger share of buyers start their shopping by asking AI first.
Get the structural answer, not just the anecdote: get your free GEO Score.