Working with motoring brands in China in the early 2000s was an education. Every stand at the auto show was surrounded by models. Vinyl outfits, fixed smiles, each one angled toward the crowd so that every elevation presented a distinctive asset.
It was a confronting scene for a marketer, and a clarifying one. The category behaviour was obvious, which meant the way to stand out was obvious too. Send the models away. Be the one stand a buyer could walk up to and actually see the car.
More than twenty years later, every brand is surrounded by models again. This time sending them away is not the move, and the category behaviour has never looked less clear. The models are large language ones now, they are not decorating your stand, and they are deciding whether anyone hears your name at all.
So ask one. Ask it a question your category owns. Who are the best independent brand strategists in Sydney. Which firms do positioning work for ASX-listed lenders. The model will answer without hesitation. It will name names. The only question that matters is whether one of them is yours.
If it is not, the reflex is to assume the work is not strong enough or the profile is not big enough. Usually that is not it. Usually the model simply cannot tell what you are.
From findable to legible
For twenty years the unit of visibility was the page. You optimised pages, you ranked, and a page could climb the results without anyone, human or machine, being certain what sat behind it. Ranking was a popularity contest, and popularity could stand in for understanding.
Models do not work that way. A model does not retrieve a page and hand it over. It assembles an answer from what it understands, and it can only understand entities it can resolve. A name it can pin to a thing. A thing it can place in the world. Claims it can lift without guessing. Search rewarded pages. Models reward entities. That is the whole shift, and most brands have not noticed it has happened.
This is the move from being findable to being legible. They are not the same, and the gap between them is where good brands quietly disappear.
A shop that outwrites the machines, illegible to the machines
I looked at a copywriting agency recently. Good shop. Real clients, the kind you would recognise on sight. Its entire pitch is that human writing beats the slop pouring out of the data centres. Sharp, funny, proudly, publicly anti-AI.
It is close to invisible to AI.
The site renders cleanly, so crawlers get the copy. That is where the good news ends. The brand name is also a common noun, so a model has no way to know whether a query is about the agency or the thing on a hot dog. The founder appears only by first name, with nothing to separate him from every other person alive who shares it. There is no structured data, so the facts a model would actually want, the clients, the location, the proof, sit on the page as unlinked decoration rather than stated relationships. And there is barely any of it. A handful of pages. Almost no surface to cite.
So when a buyer asks a model for a good copywriter, the shop that built its name on outwriting the machines never comes up, because the machines cannot read it. The irony is total. It is also the most common failure I see, and it has almost nothing to do with the quality of the work.
A brand problem in a technical costume
Here is the part that most visibility advice misses. This is not a technical problem with a technical fix. It is a brand problem wearing a technical costume.
The reasons a brand is illegible to a model are usually the same instincts that make it good to a person. Voice over fact. Implication over statement. The confidence to leave things unsaid because a human will fill the gap. A model does not fill the gap. It has no gap to fill. It reads what is on the page, and if what is on the page is all tone and no substance, it concludes there is nothing to say about you, and says nothing.
Distinctiveness used to be the whole game. Be different, be memorable, own a feeling. That still matters, and it still works, for the humans. But there is now a second reader in the room. It mediates more of your discovery than you think, it does not respond to feeling, and it responds instead to resolvability. Being legible to the machine is now part of being distinctive, because a distinctiveness nobody can retrieve is a distinctiveness nobody encounters.
The same discipline, pointed outward
Earlier this year I wrote Brand Inside the Machine, and the framework underneath it, Brand Sentinel. That piece pointed the discipline inward. Encoding your brand judgement into the AI tools your own people use, so coherence travels with the work instead of trailing behind it.
This is the same discipline pointed the other way.
Encode, the first layer, was about making your strategy precise enough for a machine to reason from. Turn the unsaid into the stated. Make the non-negotiables binary. That instruction was written for your internal systems. It applies just as hard to the open models now describing you to people you will never meet. The same vagueness that lets drift creep into your own tools is the vagueness that makes you unresolvable to everyone else's.
Encode inward so your tools stay true. Encode outward so the world's tools can read you at all. One discipline, two directions.
What this actually means in practice
I will keep this short, because the fixes are far less interesting than the shift in thinking that should precede them.
Make your entity resolvable. A name a model can pin, a founder it can place, links that tie you to the rest of your footprint. Mark up the facts. State the clients, the location, and the proof as structured relationships rather than leaving them as floating text. Write at least one plainly factual passage somewhere, a paragraph a model can quote without the jokes getting in the way. And give it enough to work with. A thin site is a thin answer.
None of that costs you your voice. The voice is for the humans. The facts are for the machine. A brand that cannot hold both is going to lose the half of its audience that now arrives through a model, and it will never see them leave, because they were never visible in the first place.
The choice you actually have
The machines are already describing your brand. Right now, in conversations you are not part of, to people deciding whether to call you. You do not get to opt out of that. The only choice you have is whether the model is describing you, or describing its best guess at you.
Pick one.
