Your Brand Is Being Shortlisted by a Machine. The Fix Is the Oldest Playbook There Is

Your customers are already asking AI what to buy, and the engines recommend just three brands per category. I analysed the reasons behind 450+ AI brand recommendations to work out how to win, and it is not a new playbook.

Somewhere today, one of your customers asked ChatGPT what to buy instead of searching Google. The answer named three brands. If yours was not one of them, you were invisible at the exact moment the decision was made, and the engines remember their favourites.

I have watched this film before. I managed some of the biggest FMCG brands in the world at Johnson & Johnson, Beiersdorf and Unilever, then spent the last ten years inside the platforms reshaping how those brands go to market, at Meta and Pinterest. When social media arrived, those of us inside the tech giants were telling CMOs to move to vertical video, build for sound off and reverse the story arc. It was a big ask, and companies took years to act. We are seeing the same wave with AI now, and the same lag, except this time it's a tsunami.

The good news is that if you have spent your career learning how business works, a good product, a fair price, real distribution and an earned reputation, then the way to win in AI is not to rip up the rule book. It is to re-engage those traditional muscles and make them legible to a machine.

The window is closing

AI adoption among consumers is already large and accelerating. Almost half of UK consumers, 47%, now say they are likely to turn to a generative AI tool like ChatGPT to research a purchase, up nine points in a single year (Attest, 2025), and ChatGPT itself pulled 1.8 billion UK visits in the first eight months of 2025, roughly five times the 368 million it took in the same period of 2024 (Ofcom, Online Nation 2025).

Despite the facts, most businesses still treat AI as a way to write emails faster, not as the place consumers now go to discover and choose brands. And the discovery layer is smaller. On Google you get ten brands, ten blue links. In the AI engines we are seeing only three recommended. If I ran a brand today that would terrify me. Even worse, the engines have memories baked in, so if you are not one of those chosen three today, it will be harder still to become one as the models update along with their memories.

Who is winning on the AI digital shelf?

I built the AI Choice Audit to answer this question, capturing brand recommendations across six engines and seven UK FMCG categories, over 450 answers, and I analysed the reason behind every one. AI is converging on a handful of players per category. In skincare, CeraVe and La Roche-Posay win, both L'Oréal brands, while Nivea and Neutrogena barely register. In coffee, Nestlé is nowhere, while Lavazza and Illy lead on heritage, because people asking about coffee are asking about good beans and good roasteries. The winners are the brands that did the fundamentals- Product, Price, Place, Promotion- and made sure they had a digital wrapper.

The machines reward the four Ps

 

Chart 1 - 4 Ps

Product shows up in 96% of all answers. The engine reads product listings like a spec sheet: what is in the formulation, what it is for, who it suits, the exact active ingredient for the exact problem. Your product pages have to cover every base.

Place is cited in 42% of answers, and the single most common reason in the whole study is simply that you can buy it in the UK. The engine wants to know you are purchasable before it will put your name forward.

Price is there too, cited in 31% of answers: the budget pick or the premium one, because the AI almost always slots a brand into a tier.

Promotion is the one that really interests me, because it is not the type of promotion you would think. It is not a clever campaign. It is an expert vouching for you, a credible source citing you, and the moment a category touches health, the machine reaches for a white coat before it reaches for a brand. In pet care, that endorsement turns up in 88% of answers. So, it is worth thinking about how you craft your campaigns and the role of powerful claims spoken by experts in your story.

 “PR matters again, clinical testing matters again, long-form and craft matter again. So welcome home, PR and storytelling, but bring structured data with you.”

Why AI recommends the brands it does

Chart 2 - 8 Reasons

SEO is not GEO

This is the misstep I am watching companies make, lifting their SEO strategy and applying it straight onto the AI engines. That is only half the story. Traditional SEO optimises a page to rank in a list. Generative engines do not rank pages; they name a single pick, then justify it with a reason.

And most of those reasons are facts the model absorbed from third-party, earned, trusted sources, not from your website. In the audit, 65% of answers leaned on an earned signal, an expert endorsement, an independent lab test, or a certification. The vet recommends you, the lab certifies you, the journalist cites you, the retailer stocks you. If that sounds familiar, it should. That is PR: professional and medical marketing, distribution, the earned half of marketing. It is your brand story. AI describes your brand; it does not just link to it. It is not SEO, and it is certainly not the optimisation trick a wave of "AEO" and "GEO" agencies are about to sell you.

 “Even if you get GEO right, it will only get you found. It is your marketing that will get you chosen.”

There is no single 'optimise for AI' brief

The six engines tested do not reason the same way. ChatGPT checks whether you are actually buyable in 71% of its answers and looks for an expert endorsement in nearly half. Google AI Overview thinks like a retailer, with availability present in 59% of its answers. Gemini is the opposite, the purest product-rationalist, raising availability in just 16% of its answers and leaning hardest on the formulation. So, the same brand needs different briefs. A brilliant formulation with poor distribution loses ChatGPT and Google AI Overview but may still show up in Gemini.

Chart 3 - Engine House Styles

The cost of waiting

I saw executives do nothing for a long time when social media started to scale, barely believing that their customers would look at Instagram instead of Vogue. This time you cannot afford to do nothing. The engines' memories harden with every model update, and the engines are settling on their three brands per category. If you are not showing up on the AI shelf today, you have a problem you need to fix right now. The brands that move now get written into the engine preferences. The ones that wait will be trying to break into a list that has already been decided.

This is the most modern marketing challenge I have come across, and the answer is the most traditional thing we know how to do. Build a genuinely good product, earn real distribution, make it visible, and price it properly. And put the money back into the reputation work the machine actually reads: the experts, the labs, the certifiers, the press, the trade. Not the campaign that persuades a shopper who is no longer making the shortlist.

The brand still has to be good. It just has to be good in a way a machine can read. And the way you make it readable turns out to be the oldest playbook there is.

Naureen Mohammed is a fractional CMO for CPG businesses. She ran the AI Choice Audit across ten categories and six engines. If you want to know what the machines are saying about your brand and what to do about it, get in touch at info@fractional-execs.co.uk