The news: Just weeks after walking back Instant Checkout, OpenAI is making another push into ecommerce with a shopping experience designed to simplify product discovery and comparison.

While the announcement was light on specifics about how the new functionality differs from existing features, OpenAI says the tool allows users to search by uploading images or describing products with constraints like budget and preferences. ChatGPT then presents side-by-side options with key details—including price, reviews, and features—to help users compare items without leaving the platform.

The feature is rolling out this week to all ChatGPT users across Free, Go, Plus, and Pro tiers.

Zooming in: OpenAI said merchants can use its Agentic Commerce Protocol (ACP) to share product feeds and promotions, ensuring their offerings are accurately represented within the platform. Retailers including Target, Sephora, Nordstrom, Lowe’s, Best Buy, Home Depot, and Wayfair have already integrated.

At the same time, merchants can build custom apps within ChatGPT to retain greater control over the customer experience and transactions, similar to integrations from Instacart and Target.

The context: The update follows OpenAI’s quick retreat from Instant Checkout, which allowed users to purchase directly within ChatGPT. That move highlighted major hurdles for agentic commerce.

Building a full marketplace requires complex infrastructure, including real-time data and inventory syncing, payments, fraud prevention, and dispute handling.

At the same time, adoption remains a challenge as retailers are wary of losing control, and many consumers are not yet ready to trust AI to complete transactions.

Implications for retailers and brands: OpenAI is clearly committed to carving out a role in ecommerce, but it remains unclear whether this new approach addresses a meaningful enough pain point to shift entrenched behavior. That challenge is evident in current usage: Just 16.7% of US AI users start product searches with an AI assistant, per a recent EMARKETER/Publicis survey.

It’s also uncertain whether ChatGPT’s results are reliable enough to drive that change—OpenAI acknowledged in November that up to 63% of product mentions in ChatGPT search results contained inaccuracies. Until the experience meaningfully improves on the status quo, adoption will likely remain limited.