The second point matters more. If people believed the answer completely, a brand would only need to be recommended. Because they check, a recommendation opens the decision and the sale depends on what the shopper finds next. I think what they are looking for is trust.

For a long time, fame has been how a brand earned a premium, because being known did most of the work of selling. If both the AI and the shopper check the evidence before they back a brand, fame stops being enough on its own. I think brands that are believed will hold their price, and that investors will start to value them that way.

Your brand now gets checked twice

Start with what happens inside the AI answer. Trustpilot paid Seer Interactive to study 804,491 AI answers about 1,926 US brands in March 2026. Review sites were the second most cited type of source. As a shopper moves from first hearing about a product to deciding, the share of citations going to review sites rises 16 times. The closer someone is to buying, the more the AI leans on what other customers have said.

The same study split brands by how they treat their review profile. Brands with no profile were cited 1% of the time. Brands that had claimed their profile were cited 54% of the time, and brands that had actively looked after it were cited 75% of the time. The study shows a link between the two. It does not prove that profiles cause citations, and Trustpilot has something to sell, but the difference is large enough to take seriously.

The second check happens in the shopper’s own mind. In one survey, 89.9% of people who shop with AI said they verify before buying, and 36.7% go to reviews first. In another, 98% check a brand they do not know after an AI recommends it. A third found that only 35.4% trust AI recommendations completely or mostly.

The AI reads the reviews and then the person reads them again, so your brand’s trust is tested twice before a sale happens.

Fame is high and trust is flat

To see how big brands measure up, I scored ten large US footwear brands with the ATTENT10N Zero-Click Score in October 2026. Each gets two marks out of 100. Demand is how much people want the brand. Trust is how far people believe it.

Demand runs from 38 for Converse to 98 for Nike. Trust only runs from 44 to 62, and no brand gets above 62. Seven of the ten sit within two points of each other, between 60 and 62. Among those seven, demand still ranges from 69 to 98.

These brands compete hard to be wanted and hardly at all to be believed, and none has pulled ahead on trust.

The gap between the two marks shows this most clearly.

Nike has the widest gap, at 37 points, then Hoka at 26 and Asics at 24. Converse is the only brand where trust is ahead of demand. Part of this is arithmetic, because when trust barely moves the most wanted brands will always show the biggest gaps. Even so, fame is running well ahead of belief at almost every brand.

The same ten brands show one more thing. How often AI names a brand tracks demand closely, so today it mostly finds the famous brands first. With only ten brands I cannot say whether it follows fame or trust. I am watching to see whether that changes once AI weighs trust more heavily.

Reviews are where a smaller brand can start

Most brands cannot outspend Nike on fame, so I think the double check on trust is good news for smaller firms. Both checks look at the same thing, which is what customers say about you, and a smaller brand can shape that. Asking happy customers for detailed reviews is something any team can start this week. I used our own audit data to test whether that effort would show up in AI answers.

First, AI clearly reads customer reviews. In answers that named Holland & Barrett, AI cited review sites 148 times, and about 109 of those were Trustpilot pages full of customer reviews.

Second, when I compared different brands, neither star ratings nor review detail told me who gets named. I tested 27 brands across six categories, using about 2,350 Google and Trustpilot reviews. I measured detail in two rough ways, by how long the reviews are and how often they mention a situation, such as “for my night shifts” or “for a wedding”. None of the measures lined up with who gets named, and I think the main reason is fame. Nike and New Balance get named on most questions whatever their reviews say, which is what the footwear scores would predict.

Third, when I compared each brand only with itself, the picture leaned towards detail. On narrow, specific questions, I checked whether a brand was named more often when its own reviews talked about the situation in the question, such as running. Of 25 brands, 16 were named more often and 9 less often. Holland & Barrett was named on 89% of those questions, against 67% where its reviews were silent on the situation.

This result is a lean rather than a finding, and the note at the end explains why, but it points the same way as my own view. I think a review that says “finally sleeping through my night shifts” gives an AI something to match against “best magnesium for shift workers”. A review that says “great product, five stars” gives it nothing to match, and gives a shopper little to believe.

My advice is to ask customers what they used the product for, rather than for a star rating. A prompt like “Tell us what you used it for and what changed” costs almost nothing, and it serves both checks, because the AI can match the words and the shopper can believe them.

What I would do on Monday

If trust is the asset, then building it is mostly ordinary brand building, done with the new checkers in mind. I have set out five actions. For each one I say what to do, why the research says it works, and what it does for the business. I use the four things an owner or investor cares about most: revenue growth, pricing power, the cost of winning a customer, and whether the brand can last.

1. Ask customers to describe the situation they bought for.

What to do: after a purchase, ask people what they used it for and what changed. Put the question in the email, the packaging and the follow-up.

Why it works: Byron Sharp and the Ehrenberg-Bass Institute found that brands grow by being linked in memory to many buying situations, which they call category entry points. A review that names a situation builds that link in public, for the AI and for the next shopper.

What it does: it lowers the cost of winning a customer, because you are recommended for more situations without paying for each one.

2. Make your facts match everywhere you appear.

What to do: check that your product details, claims and ratings tell the same story on your own site, on retailer pages and on review sites. Fix any gaps. Claim and look after your review profiles.

Why it works: Sharp’s work calls this physical availability, which means being easy to find and easy to buy wherever the shopper looks. The research on review profiles above suggests that a looked-after profile gets cited far more often. Consistency also makes you recognisable, and recognition is what lets a brand come to mind.

What it does: it supports revenue growth by reaching occasional buyers, who make up most of any brand’s customers, on the pages they actually use.

3. Publish what only you know.

What to do: share findings from your own data that nobody else has. A footwear brand might share what years of fit, returns and wear data reveal. Put your name on it and keep it accurate.

Why it works: in Sharp’s terms, you are creating something distinctive that belongs to you and cannot be copied from a competitor’s website. An AI also has more reason to quote a source that holds original facts.

What it does: it makes the brand harder to replace, which makes demand more durable and more defensible to a buyer of the business.

4. Put real people behind the brand.

What to do: use founders, athletes, store teams and communities as the recognisable face of the brand, and keep them consistent over years. Do not change the format every season.

Why it works: distinctive brand cues work through repetition, and a familiar person is a cue that survives when a machine summarises you in a sentence. Shoppers also tend to believe a person more readily than a logo.

What it does: it builds the belief that supports pricing power. A shopper has less reason to hunt for a cheaper option when they believe the brand, and a trusted face helps with that. I have not measured this effect, so treat it as my reasoning.

5. Keep funding the long term, and measure trust as well as sales.

What to do: protect your brand building budget from the pull of short-term sales activity, and track trust and review quality alongside revenue.

Why it works: Les Binet and Peter Field studied hundreds of effective campaigns for the IPA. They found that brand building pays back over years and reduces price sensitivity, while activation delivers sales now. Their rule of thumb is roughly 60% of budget on brand building and 40% on activation, though it varies by category.

What it does: it protects pricing power and lowers the cost of growth over time.

I would not treat these as a sequence, because they support each other and trust is the test that runs through all five. Proof counts only if it is believed, original data counts only if its source is credible, and a face helps only if people believe a real person stands behind it.

The established laws of brand growth still apply, but they are now tested in new places, and an AI is watching as well as the shopper.

Sources: Trustpilot and Seer Interactive, March 2026; Product.ai; Idea Grove; Alchemer; Byron Sharp, How Brands Grow (2010); Ehrenberg-Bass Institute research on category entry points and physical availability; Les Binet and Peter Field, The Long and the Short of It (IPA, 2013); ATTENT10N Zero-Click Score, ten US footwear brands, October 2026; ATTENT10N audit data, 27 brands across six categories, October 2026.

The taste-as-moat claim has become consensus in 2026.

CMOs, creative directors, founders with vibe coded products - all circling the same argument: AI cannot replicate taste, and if you have it, you w…

Read more

In 2021, a man told me that Bored Ape Yacht Club was worth more than LVMH. I laughed. I understood the excitement of being early to something. I also understood what it actually takes to build a bran…

Read more

You know that feeling when you’ve read an entire article and absorbed nothing?

When you’ve scrolled for forty minutes and can’t name a single thing you saw?

When someone asks what you think about som…

Read more

Read more

In 1962, Rolls-Royce figured something out that most businesses are exploring now….

Their Viper engines powered business jets, but reliability was poor and airlines were frustrated. The standard fix w…

Read more

Emad Mostaque is the founder of Stability AI, the company behind Stable Diffusion. Before that, he was a macro hedge fund manager whose job was finding cracks in the global financial system and betti…

Read more

….and those who extract try to increase your productivity...

This cycle is far from ending…. what happens next?

Read more

Luxembourg pays $14,333 per child per year in direct benefits. The most generous programme in the developed world. Their fertility rate? Low. Below 1.3 children per woman.

Read more

A16z recently announced media services for portfolio companies. Timeline takeovers. Embedded storytellers. Podcast access. Fellowship programmes. On the surface: content marketing with better product…

Read more

In Parts 1 and 2, we covered how feeds evolved from social graphs to semantic systems:

Era 1 (Covered in Part 1): Social graphs → Semantic feeds
Posts go out, algo…

Read more

Perplexity doesn’t show trending articles. It answers your specific question with relevant sources.

Read more

Read more

People are losing ( enjoying ) hours making videos of their friends as medieval knights, their …

Read more

Songs are inc…

Read more

We’re not burnt out from too much content, we’re burnt out from too much performance.

Read more

Our cu…

Read more

"Crypto-enabled" doesn't just…

Read more

I've been thinking about this since I started focusing more ( with my team ) on GEO - the new world of AI visibility 👀.

Read more

While everyone debates the technical merits of blockchain scalability and regulatory frameworks, the sm…

Read more