OpenAI Introduces Shopping Research in ChatGPT

OpenAI has introduced shopping research, a new feature in ChatGPT designed to automate the process of product discovery and comparison. This tool transforms product search into an interactive conversation, synthesizing information from across the web to create personalized buyer's guides.

Deep Product Discovery and Personalized Guides

Shopping research is designed for complex decision-making where users need to understand tradeoffs, constraints, and detailed comparisons. Unlike standard ChatGPT responses for simple queries, shopping research takes several minutes to conduct deep research across the internet, reviewing quality sources to deliver a comprehensive guide.

Key capabilities include:

  • Interactive Refinement: The tool asks clarifying questions regarding budget, target audience, and specific feature preferences to narrow down options.
  • Personalization: By leveraging ChatGPT memory and past conversations, the tool can tailor recommendations to a user's specific interests (e.g., factoring in gaming preferences when searching for a laptop).
  • Real-time Feedback: Users can guide the research process by marking products as "Not interested" or "More like this," allowing the model to adapt its results in real time.
  • Category Strength: The feature performs particularly well in detail-heavy categories such as electronics, beauty, home and garden, kitchen and appliances, and sports and outdoor.

Technical Architecture and Model Training

Shopping research is powered by a version of GPT-5 mini, which was post-trained on GPT-5-Thinking-mini using reinforcement learning specifically for shopping tasks.

To ensure accuracy and quality, OpenAI developed a new evaluation framework consisting of difficult product discovery queries with multiple constraints. The primary metric for success is product accuracy, defined as the percentage of products in each response that meet the user's specific requirements, such as price, color, material, and technical specifications.

Integration and Availability

Shopping research is rolling out on mobile and web for logged-in users across Free, Go, Plus, and Pro plans. To support holiday shopping, OpenAI is providing nearly unlimited usage for all plans through the holiday season.

Integration points include:

  • Manual Activation: Users can start shopping research by asking a shopping question (where ChatGPT may suggest the tool automatically) or by selecting "shopping research" from the (+) menu.
  • ChatGPT Pulse: For Pro users, shopping research is integrated into ChatGPT Pulse, which can proactively suggest personalized buyer's guides based on past conversations.
  • Checkout Path: Currently, users click through to retailer sites to complete purchases. OpenAI noted that direct purchasing will be available in the future for merchants participating in Instant Checkout.

Transparency, Trust, and Limitations

OpenAI states that results are organic and based on publicly available retail sites, citing sources and avoiding low-quality or spammy content. User chats are not shared with retailers.

Despite the improvements in citation accuracy over other models, OpenAI acknowledges certain limitations:

  • Accuracy Risks: The model may still make mistakes regarding real-time product details such as price and availability.
  • Verification: Users are encouraged to visit merchant sites directly to verify the most accurate and current details.

Sources