OpenAI is turning ChatGPT into a personal shopping assistant with the launch of a new “shopping research” experience that builds full buyer’s guides instead of quick one‑line answers. The feature, powered by a specialized version of GPT‑5 mini, is rolling out on web and mobile for logged‑in users on Free, Go, Plus, and Pro plans, with nearly unlimited usage through the holidays.
What shopping research does
Shopping research is designed for deeper decisions like comparing laptops, picking strollers, or finding gifts, not just checking a single price or feature. Instead of sending you to multiple sites, ChatGPT asks clarifying questions about your budget, preferences, and constraints, then searches across the internet to surface options, trade‑offs, and up‑to‑date details in one place.
The result is a personalized buyer’s guide that highlights top products, key differences and reasons to pick one option over another. OpenAI says this works especially well in detail‑heavy categories such as electronics, beauty, home and garden, kitchen and appliances, and sports and outdoor gear.
How the new flow works
To start, you ask a shopping‑type question in ChatGPT—anything from “help me choose a gaming laptop under $1,000” to “I need a gift for my four‑year‑old niece who loves art.” ChatGPT then switches into the shopping research interface, where it may ask follow‑ups about who the product is for, how it will be used, and what you care about most.
In the background, the model pulls live information such as prices, availability, specs, reviews, and images from retail sites and other quality sources. As options appear, you can mark items as “Not interested” or “More like this,” and the recommendations update in real time based on your feedback.
Personalized recommendations with memory
If you have ChatGPT memory turned on, shopping research can factor in what it already knows about you from previous conversations. For example, if you’ve discussed being into gaming, that context can be used when helping you find a new laptop or accessories.
The new experience also appears inside ChatGPT Pulse for Pro users. Pulse cards can proactively surface buyer’s guides based on recent topics—for instance, suggesting e‑bike accessories if you’ve recently been talking about e‑bikes. This makes the shopping support feel more like an ongoing assistant than a one‑off search.
How it’s built and how it cites
Under the hood, shopping research runs on a version of GPT‑5 mini trained with reinforcement learning specifically for shopping tasks. The model is tuned to:
- Read and synthesize information from trusted, publicly available retail and review sites.
- Cite sources directly so you can click through to merchant pages for details or to buy.
- Avoid low‑quality or spammy sources when compiling recommendations.
OpenAI notes that chats are not shared with retailers and that results are organic, not paid placements. Merchants who want to ensure their products can appear in results can go through an allowlisting process described on the site.
Limitations and what’s next
While this specialized model performs better than previous versions at citing accurate product details, OpenAI warns it can still make mistakes on specifics like price and availability. Shoppers are encouraged to confirm final details on retailer sites before purchasing.
Looking ahead, OpenAI plans to expand shopping research to cover more categories, deepen personalization, and eventually enable direct purchases through ChatGPT for merchants participating in Instant Checkout.
