Strategy

AI Shopping Agents Reward the Famous — Until a Rival Rates 0.1 Stars Higher

Meta's Muse reached 10 million shoppers in six weeks and 29% of AI buyers say it drove their last purchase. GEO gains fade fast — question-shaping is the new edge.

By Nathan Brooks

3 min read

To win the AI search wars, stop chasing the answers and start shaping the questions
To win the AI search wars, stop chasing the answers and start shaping the questionsJESHOOTS-com / pixabay

What's News

  • More than 10 million people in the U.S. and Canada adopted Meta's Muse agent in its first six weeks; shopping is its top use.
  • 29% of AI shoppers across 11 markets said AI was the strongest influence on their last purchase, per McKinsey.
  • A BCG-reported luxury fashion group more than doubled its AI-search link rate within eight weeks of fixing structured data.
  • In a skincare study, AI favored well-known brands every time until a rival's rating rose by less than a tenth of a star.
  • 88% of U.S. teenagers own an iPhone, per Piper Sandler's spring 2025 survey.

By Nathan Brooks

More than 10 million people in the U.S. and Canada started using Meta's personal AI agent Muse within its first six weeks — and the number one thing they use it for is shopping. That single datapoint, from Fast Company's analysis of the emerging agentic-commerce market, explains why brand competition is being rewritten right now.

And it's not just big-ticket purchases. A McKinsey survey across 11 markets found that 29% of people who shop with AI — nearly one-third — said it was the strongest influence on their last purchase. Groceries are now the second most common thing people buy directly inside an AI tool, just behind clothes.

What does this change for marketers?

Historically, brands grew on a human shortcut called availability bias: for everyday choices, people reach for whatever's first to mind and first to find. AI doesn't play that game. It's more literal than it is lazy, whether you're talking ChatGPT or a Grok Bot. It favors well-known brands only when everything else is equal.

A recent study of AI skincare recommendations made the mechanics brutally clear. The models picked the well-known brand every time when the products were identical. But give a lesser-known rival a rating even less than a tenth of a star higher, and that stopped. Most of the time, the study found, AI is prepared to compare — matching what a person wants against what a brand delivers.

That has fueled generative engine optimization, or GEO — making sure your brand is the answer when someone asks AI. And it works, for now. BCG reports that a luxury fashion group fixed the structured data behind its website and, within eight weeks, the rate at which AI search linked to its site more than doubled.

Here's the catch for practitioners: GEO mostly works when you're the only one doing it. In the skincare study, when every brand used the same technique, the gains all but canceled out — and brands that sat it out disappeared from the recommendations entirely. GEO looks destined to become a cost of entry, not an advantage.

Shape the question instead of chasing the answer

The alternative the piece proposes is optimizing the criteria people bring, not just how a brand presents its attributes. Marketing has a track record here:

  • Volvo made safety the way to judge a car.
  • Nike did the same for cushioning in running shoes.
  • Dyson moved vacuums from suction power to suction that lasts.
  • Avis moved car rentals from who's biggest to who tries hardest.
  • Head & Shoulders gave people something new to look for: dandruff.

Better still, frame the question around an attribute only you have. Ask AI for the best phone for FaceTime and there's only one answer — and 88% of American teenagers already own an iPhone, according to a spring 2025 survey from the investment bank Piper Sandler. In China, where people choose skincare by ingredient, L'Oréal's Pro-Xylane became one of the most sought after, and only L'Oréal brands can use the name.

Where do you start?

Find the attribute you win on, relative to your size, and check that people and AI agree. If they disagree, shaping the question can backfire in two ways:

  • If people believe something about your brand and AI doesn't, you make the question matter and AI recommends a rival.
  • If AI believes something about your brand and people don't, AI recommends you, but nobody asks.

The case study here is Dr. Whitney Bowe, the New York dermatologist who coined "skin cycling" on TikTok. She gave people a new question to ask, and retailers like Boots published guides on what to buy for it. Then she launched her own brand, with a skin-cycling kit. She shaped the question, then built the brand to answer it.

The practical takeaway for every brief going forward: two new questions belong on the agenda — what attribute do we win on, and what question do we want people to ask? As GEO commoditizes, the brands that shape the question won't have to chase the answer.

Original: protect.checkpoint.com

Share this article:

More from Nathan Brooks

Nathan Brooks

Show full bio

News editor covering regulation, policy and their impact on entrepreneurship, startups and venture funding at Business Bearings.

2,611 articles

Related articles

« Previous article