Yes, in a limited way. Google’s AI Mode can now complete a checkout at eligible U.S. retailers using Google Pay, without the shopper leaving the conversation. Anthropic separately released an open source blueprint that lets any company build a shopping agent capable of searching a catalog, comparing options, and building a cart. Neither is available everywhere yet, but both point in the same direction.
What Did Google Actually Announce?
Google’s AI Mode in Search now lets shoppers check out from eligible U.S. retailers using Google Pay, drawing on credentials stored in Google Wallet, with PayPal support coming soon. The feature works inside the same conversation where someone was researching a purchase, rather than sending them to a separate checkout page. Google has said it plans to expand the feature to more retailers and regions over the coming months, but has not published a specific date for that expansion. A separate feature, rolled out earlier in the summer, lets users connect apps like Instacart so AI Mode can build a grocery cart on their behalf. The two features shipped at different times but move in the same direction, from AI Mode describing options to AI Mode acting on them.
What Did Anthropic Release, and Is It the Same Thing?
Not exactly. On September 3, Anthropic released Claude Commerce Agents, an open source blueprint under an Apache 2.0 license. It is not a live shopping product. It is scaffolding that any company can use to build a shopping agent or a merchant agent, including an agent loop, a catalog tool layer, approval gates, and evaluation tools that a team would otherwise have to build from scratch. The shopping agent it describes can search a catalog, handle requests involving multiple items, compare options, build a cart, and answer questions about an order or a return, all inside one conversation.
Is This Already Happening at Scale?
No, and this is the part worth being precise about. Google’s checkout capability is limited to eligible retailers, and Google has not announced when it will reach more of them. Anthropic’s blueprint is a starting point for developers, not evidence that shopping agents built on it are already live and being used widely. Both announcements describe infrastructure landing, not adoption that has already happened. Anyone citing either one as proof that AI now routinely buys things on people’s behalf is overstating what has actually been confirmed.
Why Does This Matter If It Isn’t Fully Rolled Out Yet?
Because it changes what the underlying question is. For the past several years, the operating question for AI visibility has been whether a brand gets mentioned inside an answer from ChatGPT, Claude, Gemini, or Perplexity. That question does not go away, but it stops being the only one that matters once a recommendation can end in a completed transaction rather than a sentence. A system that recommends a competitor is visible. A system that quietly does not select a brand, inside a transaction it completed on someone else’s behalf, is not something a brand can see happening using the tools most companies use today.
What Should a Brand Actually Do This Week?
Start by separating what is confirmed from what is implied. A brand does not need to build an AI checkout flow this week. It does need to check whether a system like ChatGPT, Claude, Gemini, or Perplexity can currently determine its price, its availability, and its actual product attributes accurately, because those are the inputs a system would need before it could act on a recommendation involving that brand at all, regardless of when checkout features reach a given retailer.
The Underlying Question Does Not Change
Retail is the easiest place to see this shift because the action ends in something concrete, a completed cart. The same underlying question, whether a system trusts a brand enough to act on a recommendation rather than only state one, will show up in other places too, including how AI systems shortlist software, compare vendors, or assemble a recommendation for a service. The commercial action changes depending on the category. The underlying question, which brand a system trusts enough to advance, does not.
FAQs
Does Google’s AI Mode checkout work with every retailer? No. It is currently limited to eligible U.S. retailers, and Google has said it plans to expand but has not announced a specific date for wider availability.
Is Anthropic’s Claude Commerce Agents a shopping app people can use today? No. It is an open source blueprint that companies can build on to create their own shopping or merchant agents, not a consumer shopping product by itself.
Did Google and Anthropic announce these features together? No. They are separate announcements from separate companies, made within days of each other, which is part of why the direction is notable.
Does this mean AI visibility measurement is no longer useful? No. Knowing why a system trusts one source over another before it acts is still necessary, and arguably matters more once the action attached to that trust gets faster.
Can a brand test whether AI would complete a purchase from them right now? Not reliably. The answer would depend on the specific model, account, integrations, and country involved, which measures the rollout rather than the brand’s actual readiness.
What can a brand check instead? Whether an AI system can accurately determine the brand’s current price, real availability, and actual product attributes, since those are the inputs any future purchase decision would depend on.
ABOUT AXIS SUITE: Axis Suite is the AI Recommendation Intelligence platform built by TrendAxis. It measures whether AI systems like ChatGPT, Claude, Gemini, and Perplexity mention, cite, recommend, and ultimately choose a brand, and why. Learn more at axissuite.ai or explore the sourced findings behind these claims in the Proof Center.