OpenAI rolls out shopping research tool in ChatGPT to support product discovery

New DeOpenAI has introduced a new shopping research feature in ChatGPT, offering users a structured way to compare products, review options, and receive personalised buyer’s guides. The feature is now being rolled out on mobile and web for logged-in users across Free, Go, Plus, and Pro plans.

According to the company, the tool is designed to handle more detailed product decisions by asking clarifying questions, reviewing information from trusted sites, and using ChatGPT’s memory (when enabled) to factor in user preferences from past conversations.

OpenAI stated that the tool consolidates product discovery into a conversational interface, particularly for categories such as electronics, beauty, home and garden, appliances, and sports and outdoor equipment. 

For routine queries such as checking specifications or prices, standard ChatGPT responses will continue to be available, while the new feature provides more extensive research for comparisons or constraint-based decisions.

The feature can also be used to find product alternatives, compare items side by side, explore deals, or identify gifts. Users can guide the research by marking products as “Not interested” or requesting “More like this,” allowing the system to refine suggestions in real time. 

Once the research process concludes, ChatGPT generates a buyer’s guide summarising options, specifications, trade-offs, and links to retailer sites.

OpenAI said the tool is also being integrated into ChatGPT Pulse, available to Pro users, where it may surface personalised buyer’s guides based on recent discussions. The company emphasised that shopping conversations are not shared with retailers. Results are described as organic and based on publicly available retail sources, with options curated from sites that meet its quality criteria.

The system is powered by a version of GPT-5 mini trained with reinforcement learning for shopping-related tasks, focusing on accuracy, sourcing, and synthesis across multiple pages. OpenAI noted that while the model performs better than previous versions in evaluating product information, inaccuracies may still occur, including in areas such as pricing and stock levels.

OpenAI said the shopping research feature will continue evolving as ChatGPT learns more about user preferences and expands its coverage across additional categories.