Remember when asking an AI assistant to buy something felt like a party trick that didn’t quite work? You’d type “order me a phone charger” and get back a cheerful paragraph explaining what a phone charger is, maybe a list of options, and then… nothing. You still had to open a new tab, search the product yourself, and tap through five screens. The assistant was a very articulate signpost.
That gap is what Google is quietly closing in India right now. The company has started testing a feature that lets people buy products from Flipkart, the Walmart-owned retailer, directly through Gemini and AI Mode in Search. It’s a limited test covering select products and select users, with a broader rollout planned for later in October.
What actually changes for a shopper
The short version: you ask, and the AI takes you to the finish line instead of the starting line.
If you’re using AI Mode in Search and you’re looking for, say, a pair of running shoes, the flow can now carry you straight to Flipkart checkout. You’re not bouncing between a chat window, a search results page, and a shopping app. The conversation and the transaction sit closer together.
One detail matters more than it looks like it does, and I want to flag it because it’s the part most coverage glides past. The test uses Flipkart’s own checkout system, not a Google-hosted one. Your payment, your address, your order history — that stays inside Flipkart’s machinery. Google is handling the discovery and the handoff.
Why the checkout detail is the whole story
For anyone trying to understand where AI agents are headed, this distinction is the thing to hold onto. There are two very different futures hiding in a feature like this.
In the first future, the AI company owns the checkout. It becomes the store. Retailers turn into suppliers feeding a catalogue into someone else’s shopping mall, and whoever controls the assistant controls the relationship with the customer.
In the second future, the AI is a very capable escort. It understands what you want, figures out where to get it, and walks you to the retailer’s own door. The retailer keeps the customer relationship, the payment rails, the returns process, the loyalty program.
What’s being tested in India is the second model. That’s a meaningful signal about how these partnerships are being structured right now. Retailers are not, it seems, willing to hand over the last mile of the sale. And honestly, why would they? Checkout is where the margin lives, where the data lives, and where trust either holds or breaks.
The standards angle nobody gets excited about
Google has also framed this kind of capability around an open standard designed to let AI agents interact with retailers across the shopping journey, including checkout. Standards documents are the least thrilling reading material on earth, but they’re how this stuff either works everywhere or works nowhere.
Think of it like this. If every retailer needs a custom, hand-built integration with every AI assistant, you get a handful of big partnerships and a lot of small shops locked out. If there’s a shared protocol, the smaller retailer can plug in the same way the giant does. Whether this particular standard gets wide adoption is an open question, but the intent behind it tells you Google isn’t planning to negotiate one Flipkart-sized deal at a time.
Why India, and why now
Testing in India makes a certain kind of sense. It’s an enormous mobile-first market where a lot of people’s primary computer is a phone, and where typing long queries is often less convenient than just saying what you want. An assistant that compresses six taps into one conversation has more obvious value there than it does for someone who already has a desktop bookmark bar full of shops.
What I’d watch for as a normal person
A few practical things I’d keep an eye on if this reaches you:
- Who you’re actually buying from. If the checkout is Flipkart’s, your returns, warranty, and customer service questions go to Flipkart, not to Google.
- How results get chosen. When an assistant recommends one product instead of listing twenty, the ranking logic becomes a lot more consequential. Ask yourself why this item surfaced.
- How much you’re confirming. A good version of this shows you the price, the seller, and the total before anything is charged. Convenience should never mean surprise.
The interesting thing about this test isn’t the technology, which is mostly plumbing. It’s the shape of the deal. Two large companies worked out who owns which part of a purchase, and they landed on the retailer keeping the register. Every AI shopping partnership that follows will be measured against that line.
🕒 Published: