Amazon wants your shopping agent to be Amazon's, not Meta's

Amazon's new feature lets you text its AI agent a photo of a recipe and have the ingredients land in your cart. Retail Dive reports the company switched this on for select US shoppers on Friday, with a wider rollout to millions more promised within weeks. The convenience is real, but read the move against Amazon's recent fights and a different story emerges: this is about who owns the agent that shops Amazon, and the answer Amazon wants is Amazon.

The texting feature is the friendly face of a turf war

Over the past year Amazon has moved aggressively to keep AI shopping inside its own gates. In September it asked to be pulled from Meta's Muse experience, saying it never authorized Muse users to buy from its site. In 2025 it sued Perplexity after that company's Comet agent was placing orders on Amazon for users. On an October earnings call, CEO Andy Jassy said plainly that Amazon expects Alexa for Shopping to be the main agent people use to shop Amazon, even while it works with outside ones.

Put those together and the text-to-shop launch stops looking like a standalone product. It is the carrot alongside the stick. The lawsuits and access removals push outside agents away; a slick texting tool pulls customers toward Amazon's own. Every purchase that happens through Alexa for Shopping is a purchase that did not route through a third party, and the behavioral data stays with Amazon.

Not first, and that's the point

Text-to-shop itself is old news. Walmart shipped a Text to Shop feature back in 2022 through its No. 8 incubator. What Amazon adds is its rebranded AI agent underneath, plus Shop Direct, which surfaces products from across the web inside the same chat. The novelty is less the texting and more the strategic bet that owning the agent layer matters more than owning any single feature.

This is where the stakes sit for anyone building a shopping agent. If the pattern holds, the large retailers each wall off their catalog behind a house agent, and the independent assistants that promised to shop everywhere at once find the biggest stores locked or litigated out of reach. A universal buy-anything agent gets harder to build, not easier.

The catch is whether anyone trusts it to buy

There is an honest limit here, and it is demand, not technology. People already use AI to get product ideas and compare prices. Handing an agent your card and letting it complete a purchase is a different level of trust, and plenty of shoppers balk at it over security worries. A recipe-to-cart demo is low risk because you still review the cart; an agent that checks out on your behalf is where hesitation kicks in.

So the feature can roll out to millions and still mostly get used for the browsing parts. Amazon's ownership play only pays off if customers actually let the agent close the sale. The rollout pace will show intent, but the adoption of that final buy step is what will show whether the strategy works. Until shoppers trust an agent to spend their money, owning the agent is worth less than it looks.