When a customer asks an AI assistant what to buy, does your brand come up as an answer? New research from Georgetown and Darden, published in Harvard Business Review as “How to Get AI to Surface Your Brand” (Gale, Cian and Wathieu, June 2026), suggests many household names don’t. Across 15 retail categories and three major AI platforms, the researchers found that fewer than one in ten brands appeared consistently across systems, and that some of the world’s most recognisable names were absent entirely — while a focused player like Brooks surfaced reliably in running shoes. The reason is structural. AI assistants don’t reward awareness or storytelling; they work forward from the user’s problem to a product attribute to a brand that demonstrably delivers it. Inclusion in that reasoning chain — not sentiment — is the competitive bottleneck. The brands that get retrieved are the interpretable ones: clearly identifiable entities, with named, measurable attributes, backed by independent evidence.

That is an intellectual property strategy problem wearing a marketing costume. Look at what interpretability actually requires: entity clarity is trade mark and brand architecture discipline — consistent naming, clean ownership, sub-brands distinct enough to be retrieved on their own merits (the study found parent brands often surface only through specific products). Attribute structure is your patented and protected technology expressed as verifiable specifications rather than adjectives. The evidence base is decades of third-party validation that money can’t shortcut — which makes it a moat.

The practical move for any IP-intensive business: query the major AI platforms with your customers’ real problem language, audit whether your protected differentiators are nameable and measurable, and assign cross-functional ownership of how your brand is described everywhere it appears. This echoes two patterns we’ve written about before — that AI Isn’t Your Advantage—Your IP Strategy Is, and that Your IP Strategy Is Only as Good as Your Board’s Questions. The forward-looking takeaway: generative engine optimisation isn’t a media tactic. Brands built on defensible, well-structured, evidenced IP were always easier to explain — and it turns out machines agree.