Google Tests Flipkart Checkout in Gemini AI Mode

Google is testing direct purchases from Walmart-owned Flipkart inside Gemini and AI Mode in India. The test adds a “Buy” button to select Flipkart listings returned in conversational results, taking users to a Flipkart checkout flow without leaving the AI interface. TechCrunch

The scope is narrow. Access is limited to some users and a small selection of products, covering smartphones, electronics and mobile accessories. Google plans a broader rollout later in October, ahead of India’s festive shopping season.

The implementation keeps the transaction merchant-hosted. Tapping Buy invokes a Flipkart-branded checkout flow inside the AI surface. That differs from the Google-hosted checkout experience Google previously demonstrated for agentic purchases.

The plumbing is the Universal Commerce Protocol (UCP). Google introduced UCP earlier in 2026 as an open standard to let AI agents interact with retailers across the shopping journey, including checkout. The company said UCP will power a checkout feature for eligible product listings in AI Mode in Search and the Gemini app. Google

On the discovery side, Google’s AI Mode shopping experience combines Gemini capabilities with its Shopping Graph. Google Flipkart was named among the merchants working with Google to bring “agentic” shopping experiences to consumers in India.

The two companies already have a capital link. Google invested about $350 million in Flipkart in 2024 as part of a funding round led by Walmart, taking a minority stake. Walmart’s control position dates further back. Walmart confirmed a $16 billion investment in Flipkart for a 77% share in 2018. TechCrunch

The broader context here concerns where the cart lives and who owns the trust boundary. A Flipkart-branded flow inside Gemini suggests session handoff rather than platform intermediation. Product discovery, ranking and conversational refinement happen in Google’s interface, while pricing, payment credentials, order creation and post-purchase handling stay inside Flipkart’s stack. For engineers building agent integrations, that division matters. It preserves merchant logic for inventory, coupons, fraud checks and returns, and it avoids duplicating that logic in the assistant.

In this author’s view, the choice is pragmatic and worth watching as a template. A Google-hosted checkout offers fewer clicks and cleaner telemetry, but it asks retailers to cede the most sensitive part of the funnel. A protocol-based, merchant-hosted checkout lowers that barrier, at the cost of more complex state synchronization between two systems that must agree on identity, cart contents and payment confirmation. If UCP succeeds as an open layer, the differentiator will not be who renders the button. It will be how reliably agents and retailers can pass structured commerce context back and forth without leakage, drift or abandoned carts.

Looking at what this means for deployment, timing the broader release for October puts the integration under real concurrency and catalog churn. Festive catalogs change quickly, with variants, bundles and price shifts that stress retrieval and grounding. A limited test on phones and accessories provides a constrained SKU set with well-structured attributes before that peak. The outcome to track is operational, not conceptual: completion rates, error handling when listings change mid-conversation, and whether users accept finishing payment without switching to a separate app.