I came to System 3 through Adam Ferrier of Thinkerbell, and I love the idea so much I want to share it — along with credit where it belongs. The concept is not his: it comes from Steven D. Shaw and Gideon Nave at the Wharton School, who introduced it in January 2026.

You know Daniel Kahneman and his System 1 and System 2 framework that continues to be a massive influence in cognitive and behavioural science spheres. This being the fast, emotional brain mode and the slow, rational, methodical brain mode. Brand and marketing has spent decades learning to understand and leverage both. Shaw and Nave's argument, in Thinking—Fast, Slow, and Artificial: How AI is Reshaping Human Reasoning and the Rise of Cognitive Surrender, is that there is now a third: artificial cognition that operates outside the brain. Not a channel. Not a tool. An extension of how people actually think and decide. Ferrier has been making the same case to marketers on Black T-Shirts, and that is how the idea reached me.

I absolutely subscribe to this thinking. And I want to push the idea one step further, because there's a consequence hiding inside it that most brands haven't digested yet.

If your buyer thinks through a machine, then (maybe) the machine's opinion of you is now part of your buyer's mind…

For your entire career, brand building meant earning a place in human memory. Distinctiveness, salience, associations, trust — all of it lived in someone's mental model of you. Be remembered, be preferred, be chosen.

But System 3 doesn't live inside a consumers mental model. It lives in a machine model. And it has already formed a version of your brand assembled from whatever it could find and verify, without your packaging, your film, your art direction, or your permission. When your buyer asks it something, that's the version that answers.

I watched this happen last week. A consultancy I work with got an inbound from a Fortune 500 company — a serious, name-brand buyer with a nine-figure problem. Asked how she found them, she said: I was researching through AI. But she didn't "search". She thought, with a machine as part of the thinking, and the machine put this firm forward.

Here's the uncomfortable part. When I measured that same firm, three out of four leading models couldn't reliably identify them. One thought they were a hotel's in-house team. Another confused them with a magazine. The models were also inventing client relationships that never existed. The firm was winning through System 3 and being misrepresented by it at the same time — and they had no idea either thing was true.

Shaw and Nave measured what happens next, and it is worse than indifference. Across three preregistered experiments — 1,372 people, 9,593 trials — accuracy rose 25 points when the AI was right and fell 15 when it was wrong. Confidence went up either way, including after the errors. They call it cognitive surrender: adopting the machine's answer with minimal scrutiny. The people who trusted AI most surrendered hardest. So when a model misidentifies your firm, your buyer does not catch it. They believe it slightly more than they would have believed the truth.

That's the state of most brands right now. Not absent from the machine's mind. Present, and wrong.

This is why I think the frame matters more than the tactics.

The industry's first response to AI search has been to treat it like SEO with a new haircut — structure your content, feed the crawlers, chase the rankings. That's plumbing, and yes, you need plumbing. But it misunderstands what System 3 is doing. System 1 could be seduced. System 2 could be persuaded. System 3 needs be convinced and it convinces itself by verification. It reads what you claim, checks it against what the world says back, and decides whether you add up. If your promise doesn't survive that check, it doesn't argue with you. It just skips you and recommends someone it understands better. You'll never know it happened. Neither will your buyer.

There's an old word for what's being tested here: Trust. The confidence that you'll deliver on your promise. It's the same thing brand building has always been about. The only thing that's changed is who's doing the checking — and unlike humans, this audience never gets tired, never skims, and never gives you the benefit of the doubt on a claim it can't verify.

Ferrier's crystal ball, as usual, was pointed the right way, and Shaw and Nave have now put numbers under it: the winners of the next decade won't be the ones with the most AI. I'd add: they won't be the ones with the most content either. They'll be the brands System 3 trusts — the ones whose promise, evidence, and voice hold together when a machine takes them apart and checks.

Most brands have no idea whether that's true of them. That's the blind spot. That's what we measure.

ARA is the quarterly measure of machine trust — what AI systems believe about your brand, whether they can verify it, and whether they'll recommend you when your buyer asks. System 3 and Tri-System Theory are Shaw and Nave's; Adam Ferrier of Thinkerbell brought the idea to marketers and to me; the conclusions I've drawn from all of it are mine. Sources: the paper on SSRN, Shaw and Nave in conversation on the Humans + AI podcast, and Ferrier's own Black T-Shirts.