Should brands let consumers buy their products through checkout inside AI assistants?
Short answer
Selectively, and on terms. Checkout inside an AI assistant removes friction but can move the transaction, customer record and pricing visibility to the platform. Brands should prefer arrangements where they or an authorized retailer stay merchant of record, receive order and customer data, and know the fees. Platform models are still changing within months.
Key points
- “Checkout inside an AI assistant” covers three different models with very different consequences for the brand.
- The deciding questions are who is merchant of record, what data comes back, what it costs and whether the brand can exit.
- Many brands will reach AI checkout through retailers’ integrations before their own.
- A small, measured pilot beats both refusal and a rush to every platform.
What does checkout inside an AI assistant mean in 2026?
| Model | How it works | Examples as of October 2026 | Who keeps the customer |
|---|---|---|---|
| 1. Handoff | The assistant recommends and sends the shopper or cart to the merchant’s own checkout | OpenAI’s current approach for most merchants; Google’s “send cart to merchant site” option | Merchant |
| 2. Native checkout | The shopper pays inside the assistant; the merchant fulfills | Google’s “pay on Google” option through UCP; OpenAI’s original Instant Checkout | Shared; depends on data terms and account linking |
| 3. Agent purchase | An agent completes the purchase on the merchant’s site for the shopper | Amazon’s Buy for Me; Perplexity’s Comet | Merchant holds the order, but often has little context about the shopper |
Retailer apps inside assistants add a variation. Walmart launched an app in ChatGPT in March 2026 that supports account linking, Walmart payments and loyalty, which keeps the relationship with Walmart.
What has happened so far?
OpenAI introduced Instant Checkout and its Agentic Commerce Protocol on September 29, 2025. In March 2026 it said the first version “did not offer the level of flexibility” it wanted, and it now lets merchants use their own checkout while it focuses on discovery. Target, Sephora, Nordstrom, Lowe’s, Best Buy, The Home Depot and Wayfair supply product data through the protocol, and Shopify’s catalog feeds ChatGPT for Shopify merchants.
Google announced the Universal Commerce Protocol in January 2026 with Shopify, Etsy, Wayfair, Target and Walmart. On October 9, 2026, it opened a Merchant Center hub for a phased US rollout that lets selected merchants add checkout to eligible products in AI Mode and Gemini. Merchants choose whether to send the cart to their site, accept payment on Google or link shopper accounts.
The lesson for brands is that platform models are still changing within months. Commitments should be easy to adjust.
What are the benefits and risks?
Benefits
- Conversion. Fewer steps between recommendation and purchase. Adobe reported that AI-referred visitors converted 31% more often than other traffic in the 2025 holiday season, even with today’s handoffs.
- Visibility. Platforms may favor merchants that can complete transactions inside the experience.
- Early learning. Brands that participate early learn how agents represent and route their products.
Risks
- Customer access. Native checkout can leave the brand with an order and little relationship. See who owns the customer relationship when an AI agent buys.
- Economics. Platforms that control the transaction can introduce fees, promoted placement or preferred-merchant programs. See whether brands will pay for placement in AI shopping agents.
- Pricing and channel conflict. Side-by-side price comparison inside an assistant exposes MAP violations and retailer pricing differences immediately.
- Service and returns. Orders placed through an assistant may arrive without context, raising error and return rates. See how to handle returns from agent-generated purchases.
- Measurement. Attribution across assistant, retailer and brand gets harder. See how to attribute sales generated through AI agents.
What terms should a brand require before enabling AI checkout?
- Merchant of record: the brand or an authorized retailer, not an unknown third party.
- Data: order details, customer contact with consent and the ability to link accounts and loyalty.
- Fees: written, with notice periods for changes.
- Representation: product information sourced from the brand’s authoritative feed.
- Routing: orders limited to authorized sellers.
- Policies: returns, cancellations and disputes handled consistently with the brand’s own channels.
- Exit: the ability to modify or disconnect the integration without losing customer records.
Should brands sell directly or through retailers in AI assistants?
For brands that sell mostly through retailers, the first question is how those retailers represent the brand inside AI assistants. Target, Walmart, Best Buy and others already feed AI platforms. Brands should confirm that retailer feeds carry accurate, complete product data, and decide whether a direct integration adds enough value to justify potential channel conflict. DTC-heavy brands on Shopify or similar platforms can often participate through their commerce platform with limited custom work.
How should a brand run a pilot?
Choose a limited set of SKUs with simple purchase decisions and low return rates. Use the handoff model or a native model with account linking. Set success measures in advance: conversion, contribution margin, return rate, share of customers identified and incrementality versus existing channels. Review after one to two quarters before expanding.
What good looks like
The brand participates in AI checkout where terms keep it or its authorized retailers as merchant of record, data flows back, fees are known and integrations can be changed. Pilots report contribution and customer capture, not just orders. Management revisits the decision as platform models change.
Frequently asked follow-up questions
Does ChatGPT still offer Instant Checkout?
OpenAI changed course in March 2026. It said the first Instant Checkout did not offer the flexibility it wanted and now lets merchants use their own checkout experiences while it focuses on product discovery. Merchants can still build deeper integrations as ChatGPT apps.
Can merchants add checkout to Google AI Mode and Gemini?
Yes, in a phased US rollout that began in October 2026. Merchants build a Universal Commerce Protocol integration, apply through Google Merchant Center and, if selected, test in a sandbox before going live. Google has said Australia and Canada will follow.
Do brands need to support both OpenAI’s and Google’s protocols?
Brands should support the protocols their commerce platform makes available at low cost and focus custom work on the platforms that reach the most shoppers in their category. Supporting a protocol does not oblige a brand to accept a platform’s commercial terms.
Sources
- Digital Commerce 360, OpenAI reveals updates to its agentic commerce experience for ChatGPT, March 24, 2026.
- InfoQ, Google’s Universal Commerce Protocol, January 2026.
- Shopifreaks, Google Merchant Center UCP hub, October 9, 2026.
- Digital Commerce 360, Adobe data on generative AI holiday shopping traffic, 2025, January 13, 2026.
- Amazon, Alexa for Shopping announcement, May 13, 2026.
Related questions
- What is an agentic commerce protocol?
- What is the Universal Commerce Protocol and why does it matter to brands?
- Do consumer brands need to support multiple AI commerce protocols?
- How will AI shopping agents make money?
- Which AI agents and platforms should brands allow to sell their products?
- Agentic Commerce Protocols for Executives
The Agentic Commerce Brief
What changed in agentic commerce, and what it means for brands.
A concise weekly brief for brand executives navigating the shift from human-led shopping to agent-mediated commerce. Follow the developments that matter, the implications for brand control and governance, and the questions senior teams should be asking now.
Read The Agentic Commerce Brief on SubstackIf the signup form does not load, subscribe directly on Substack.
