Trends Podcast

It took twelve days for Amazon to lock Meta's new AI agent out of its store.

The agent is called Muse. In its first five days, seven hundred and thirty thousand people downloaded it. Meta, the company behind Facebook and Instagram, launched it in the US on the eighth of September, twenty twenty-six.

Meta says Muse "can make suggestions unprompted and act on details that person only mentioned once." You save a recipe reel on Instagram. Muse turns it into a grocery list and gets the order ready to buy.

On the twentieth of September, Amazon told its customers it blocked Muse from shopping on Amazon.

The next day Shopify opened about a million stores to it.

The opportunity I see in this fight belongs to small Shopify brands that learn to get picked by agents. It also belongs to the people who help them.

Before the block, a YouTuber who runs a channel called The Tech Girl tested Muse on her Amazon account. She asked it to reorder the sunscreen she'd bought before. Muse started a computer Meta runs for it in the cloud, opened Amazon in a browser and logged in.

It went to her order history, found the sunscreen and put it back in the cart.

Then, in her words, it went on to "proceed to check out, all on its own."

She raised the worry herself: an agent holding her Amazon password. Her answer was Muse's secure credential store, a space she said "cannot be seen even by Meta." Meta added one check: a second agent, called the Sentinel, has to approve each action Muse sends to the internet.

Muse climbed to the top free app on Apple's App Store, above ChatGPT.

A Meta spokesperson told CNBC that shopping "has quickly become one of Muse's biggest usage drivers."

Then Amazon's notice went out. Amazon said apps that buy on a customer's behalf "should operate openly and respect service provider decisions about whether or not to participate." It said Muse didn't identify itself as an agent when it shopped.

Amazon called Muse "an undisclosed third party moving through customer accounts." CNBC reported that Amazon also accused Muse of storing customer passwords and scraping account data. Muse users who gave it their Amazon login lost the agent that reordered for them.

Those may be fair complaints.

Amazon also sells a shopping agent that does what Muse does. Since May, Alexa for Shopping has checked out on other retailers' sites with a feature called Buy for Me. Alexa Plus, free with Prime in the US, has Auto Buy: you set a target price and Alexa buys the item when it drops to it.

Amazon's agent shops at other stores. Meta's agent is barred from Amazon's.

Picture the shopper Amazon wants to keep.

She types sunscreen into Amazon's search box. She scrolls past sponsored listings and lands on a product page with ads around it.

Amazon earns at each step. With Muse the agent does the searching and reads the pages.

She sees the finished order.

Ben Thompson, who writes the tech newsletter Stratechery, put it in four words: "what is scarce is volition." He means the want itself, the moment you decide you need sunscreen.

The company whose agent suggests the purchase first decides which store gets the money.

Amazon blocked Meta's agent, so on Amazon the agent that suggests her next purchase is Alexa.

The next day Shopify let Muse in.

On the twenty-first of September Shopify switched agent checkout on by default for about a million eligible US stores. A store owner who did nothing was now open to Muse. To opt out, she'd have to go into Shopify's settings, find the Meta channel and switch direct checkout off.

Shopify's founder, Tobi Lütke, said, "We're making every Shopify store agent-ready by default."

Shopify's numbers backed the move. In the first quarter of twenty twenty-six, traffic to its stores from AI tools grew eight times over the year before. Orders from AI search grew about thirteen times.

Big chains signed on with Meta too. Walmart, Gap and Sephora were in Meta's Muse announcement. Mark Zuckerberg said Meta will take a small fee on each purchase.

Now picture one of those million stores.

A brand owner sells sunscreen on Shopify. A shopper tells Muse to restock. Muse reads the product pages of a few stores that sell sunscreen and picks one.

It reads the text on each page and skips the photos. Her page left out the details the agent checks, so the sale goes to another brand. All she sees is a slow week.

Pay attention to the next sentence, because it's the most important one in this video.

An agent can only buy what it can read. Most product pages were written for people looking at photos.

OpenAI let shoppers buy inside ChatGPT, then shut that checkout down this year. The product details ChatGPT pulled from store websites, like what was in stock, delivery times and costs, kept coming back wrong or out of date.

Big brands pay to fix this.

ReFiBuy's software reads a brand's catalog the way a shopping agent does, fills in the missing product details and sends the fixed data back to the agents. Steve Madden's head of direct-to-consumer is quoted on ReFiBuy's homepage. In May ReFiBuy raised a thirteen point six million dollar seed round.

A second company, Azoma, tracks whether Alexa for Shopping, Walmart's Sparky, ChatGPT and Gemini recommend a brand's products. Then Azoma works on the product data to raise the odds they do. L'Oréal, Unilever and Mars are its customers.

Both companies sell to brands with enterprise budgets. A Shopify brand doing one million to twenty million dollars a year is too small for their sales teams. Shopify opened that brand to all agents by default.

That's the opportunity I see in Amazon blocking Meta's agent.

The buyers are Shopify brands too small for ReFiBuy and Azoma. They'd pay to get their products picked when a shopping agent buys. The opening came on the twenty-first of September, when Shopify flipped its switch. And the proof is ReFiBuy and Azoma selling the same fix to Steve Madden and L'Oréal today.

I mapped this market for the Trends dot V C report.

The plan there starts as a service a small team can run.

The team picks one brand and takes its top fifty products. It asks Muse, OpenAI's Dots and Gemini to shop for each one. The team writes down which products the agents skip and why.

Then it rewrites those product pages and runs the agents again until they pick them. The first brands are in Shopify's partner directory and in the communities where direct-to-consumer founders swap notes.

The pitch fits in one line: "See which of your products the shopping agents skip, then get them picked."

On a panel about agent shopping, Visa's Godfrey Sullivan said an agent looks across each merchant that sells what you asked for. In his words, "it brings a lot more opportunity for those smaller merchants."

Amazon kept Meta's agent out of one store. Shopify sent it into a million others, where the sale goes to the brand whose page the agent can read.

A million Shopify stores opened their checkout to Meta's agent less than two weeks after it launched. Visit trends dot V C and join more than fifty thousand founders to get the next opportunity.