When AI Becomes the Buyer

Picture a runner who needs new shoes before a half marathon. She opens Meta’s Muse app and types what she wants: under $120, good for flat feet. Muse already has a sense of her taste from years of Instagram likes. It looks at a handful of pairs, settles on one, fills in the checkout and shows her the total. She taps approve and gets on with her day.

Now picture the shoe brand that didn’t get picked. It may have spent more on ads than the winner. It may have had the better store display. None of that came up. There was no visit to its site and no abandoned cart to chase. It lost a sale without ever knowing it was in the running.

That is the shift brands are living through right now. For as long as there have been brands, the job was to persuade a person. More and more often, the first one you have to persuade is a piece of software.

It started this autumn

Meta launched Muse in the US on September 8. It can search for products, work through a checkout and pay once you approve the total (Fortune). People wanted it. Muse passed 730,000 downloads in its first five days and overtook ChatGPT as the top free app on the App Store (SiliconANGLE). Two weeks later, Shopify announced that shoppers could check out with Shop Pay inside Muse at every Shopify store (PYMNTS).

The big retailers then picked sides. Amazon blocked Muse from shopping on its site, saying the agent doesn’t identify itself as an AI (SiliconANGLE). Walmart did the opposite. In October it opened a gateway that lets approved AI agents search its shelves and buy for customers, within spending limits the customer sets (TechCrunch). When the largest retailers in the world start fighting over who holds your customer’s shopping list, it is a good sign the shopping list matters.

Startups are moving too. Instinct, founded by former Sierra researchers, has built an assistant you simply text. It books trips, clears your inbox and handles shopping, and its terms allow it to make purchases on your behalf. It is reportedly raising money at a $2.5 billion valuation (TechCrunch). Google, meanwhile, is adding a Buy button inside AI Mode and Gemini, so a shopper can purchase without ever leaving the answer (Google).

The shoppers who arrive are different now

For a while, traffic from AI assistants was easy to dismiss. It was small, and the people who came that way tended to look around and leave. Both of those things have changed.

Adobe tracks visits to US retail sites. In September, traffic from AI assistants was up 71% on a year earlier. More telling is how those visitors behave. A year ago they converted worse than everyone else. In September they converted 90% better, and each visit brought in 83% more revenue (Adobe).

Source: Adobe Analytics, 2026 · View chart full size

These shoppers aren’t browsing. The AI has already done the comparing, and they arrive ready to buy. And the biggest test is weeks away. Adobe expects AI traffic to retail sites this holiday season to more than double on Thanksgiving and nearly double on Black Friday and Cyber Monday.

Source: Adobe forecast, October 2026 · View chart full size

The habit is already mainstream. In a 2026 survey of more than 1,000 US consumers by Exploding Topics, part of Semrush, 77% said they had used AI to help them shop in the past six months, and 43% do it every week (Search Engine Land). Juniper Research expects AI agents to drive $1.5 trillion in purchases worldwide by 2030 (Juniper Research).

People are still wary of handing over the wallet. Only 31% say they’d let an agent do their shopping, and 72% wouldn’t give one their card (Fortune). In the Exploding Topics survey, the typical shopper would let an AI spend about $50 on its own. But that misses where the real decision happens. Long before anyone pays, the AI has already chosen which three or four products make the shortlist. If you’re not on it, the question of who presses buy never comes up.

A machine doesn’t shop the way we do

An agent doesn’t get excited by a campaign or feel loyal to a logo. It works from what it can find and read about you. Wharton professor Kartik Hosanagar has written about what decides whether an AI recommends a brand in the first place. It comes down to whether its crawlers can reach your site, how you show up in what the model has learned, which search index it relies on, and the words on your pages (Creative Intelligence).

And there isn’t just one of it. Each assistant draws on different sources and reaches its own conclusions, so being picked by one tells you little about the others. Shoppers spread their questions across many of them. Among people who shop with AI, 78% use ChatGPT and 58% use Gemini, and many use both (Search Engine Land). Now Meta Muse has joined them.

What this means if you’ve always sold to people

Most of what brands have built over the years was designed for human eyes: the packaging, the shelf placement, a memorable ad, the feeling someone gets when they see your logo. An agent experiences none of it. It reads your price, your sizes, your stock, your return policy and your reviews, and it decides from that. In Adobe’s latest check, the average US retail homepage scored just 63% on how much of it AI can actually read (Adobe). A product the machine can’t read is a product it can’t pick.

It also reads what other people say about you, and this is where many brands get caught out. The agent pulls from review sites, Reddit threads, YouTube videos, retailer listings and comparison articles. A stale listing with the wrong price, or a forum thread calling you overpriced, can describe you to the buyer more forcefully than your own website does.

Advertising doesn’t carry over neatly either. Meta says Muse conversations won’t be shared with its ad systems (Fortune). When an agent makes the choice, the moment an ad used to work in may simply not be there.

The hardest part is that none of this shows up in your numbers. There is no click to track and no cart to recover. A brand can lose a growing share of its sales to AI shortlists and see nothing in its dashboard, because nothing ever happened on its site.

So the first job is to see clearly: what each AI says about you, which sources it relies on, and where it gets you wrong. The second is to fix those sources, not just your own pages. The third is to test changes before you make them, because AI answers shift from day to day and a live change can take weeks to show.

How Bodhium Labs helps brands win with AI buyers

This is the work we do with brands every day, through the Bodhium Labs AEO platform and a team that works alongside yours.

We track how every major AI platform talks about your brand and your products. That includes ChatGPT, Gemini, Meta Muse, Perplexity, Claude, Grok, Copilot and Google AI Mode, along with the product cards and shopping results these assistants show. You see where you appear, how you are described and how you compare with your competitors.

We show you where the answers get you wrong and why. Every mistake traces back to a source, whether that’s a retailer listing, a review, a Reddit thread or a YouTube video, so you know exactly what to fix. We also give you regular insights across Reddit, YouTube, LinkedIn and other channels, and we monitor what is making your competitors win.

We tell you what to do first. Instead of a dashboard to interpret, you get a ranked list of opportunities, ordered by the impact each one is likely to have.

We help you make the changes. That covers shopping feeds and product pages, blogs and articles, content edits and site optimization, and the third-party and affiliate content that agents trust.

And we test before you publish. Bodhium Labs is the only company that can test content before it goes live and tell you whether it will work. You can put two drafts, or two whole content strategies, side by side and see which one AI answers favor before anything goes live. That helps you avoid costly mistakes and get to the best strategy first, instead of publishing, waiting months and guessing again.

That is why our clients see results fast. Ridge went from 0% to 46% visibility in ChatGPT for non-branded searches and showed up in recommendations within a week during its peak season. Mando grew its non-branded AI visibility from 7.8% to 28.1% (bodhiumlabs.com/geo). Our team is research-backed and measured on one thing: results for the brands we work with.

The buyer has changed

Your next customer may never visit your site. More and more often, an AI will read about you, weigh you against three others and decide in a few seconds. The brands that do well will be the ones that know what it’s reading.

If you want to know how AI sees your brand today, and what it would take to be the one it picks, talk to the Bodhium Labs team.