TheMachineAlreadyHasanOpinion

Someone is about to ask a model which running shoe to buy.

Not search for one. Ask. They will describe their arches, their weekly mileage, the calf strain that flared up in March, and they will wait for an answer that arrives as a sentence rather than a wall of blue links.

The interesting question is what happens to the £180 shoe in that moment. Does it survive the sentence, or does a £90 pair with the same foam win because the machine cannot feel the difference between them?

The tempting answer is that the premium dies. A cold engine lines up the specifications, finds the cheaper equivalent, recommends it without sentiment. No badge, no heritage, no campaign survives contact with a reader that cannot be flattered. Marketing built on feeling finally meets someone with no feelings.

It is a clean story. It is also wrong, and the spot where it breaks is the spot worth standing on.

A language model does not compute a verdict from first principles. It recalls. Everything it says about a product is a compression of the millions of sentences people have already written about that product, its category, and the kind of person who buys it. When it calls the £180 shoe a favourite among serious marathoners, no measurement is taking place. It is passing on a reputation it absorbed. The bias everyone assumes has been stripped out is, in fact, the raw material it works from.

So what happens to brand value if it doesnt evaporate?

A brand was never only decoration. It is a shortcut for trust, a way to avoid evaluating every purchase from nothing each time. Under delegation that shortcut matters more, not less. The agent acting on your behalf faces genuine uncertainty. It needs proxies for the things it cannot verify, whether the product lasts, whether the company answers when something breaks. Reputation is the proxy. The model leans on it precisely because it cannot test the thing itself.

What changes is where the contest happens. It used to take place on a shelf or a screen, in the half second when a human felt something. Now much of it happens earlier and out of sight, inside the weighting the model has already assigned before anyone types a word.

Which points at the real threat, and it is not the comparison. A brand rarely loses the comparison. It loses by never entering it. The model offers three options. If you are the fourth, your premium is irrelevant, because no one will ever see it to reject it.

We keep making the same mistake about every new technology. We assume it has rewritten human nature when it has only moved the furniture. People still want reassurance, still pay more for the thing that feels safe, still hand their judgement to a name they trust. The machine has not killed that instinct. It has inherited it, and changed the address where it gets satisfied.

There is a real cost buried in that move, and brands should be honest about it. What the machine can read, it can reward. What it cannot read, it tends to leave out. A premium backed by something legible, durability that surfaces in reviews, service that leaves a trail, expertise that other experts write about, is easier to defend now than it has ever been. A premium backed by pure mystique, by a feeling no one ever wrote down anywhere a model could reach, is suddenly exposed.

The ineffable does not compress well.

So the work is not to charm the machine. You cannot. The work is to make the reasons your premium exists legible to a reader that only knows what has been written down. Put the substance where it can be found. Earn the sentences before the question is ever asked.

Comparison was never the thing to fear, being absent from the answer is.