OpenAI wants consumers’ buying journeys to start with prompts, rather than product pages.
The technology company said Monday that it has launched an additional tool, which it calls shopping research, to help consumers with product discovery for complex or technical inquiries.
Shopping research will allow a user to input their request — for instance, “Help me find a pair of burgundy booties with a block heel and a square toe.” From there, it populates a quiz-like experience that helps it narrow down exactly what the user is seeking — it often asks questions about budget or further product attributes.
Related Articles
Once the user completes the short quiz, they are presented with an opportunity to steer the model. ChatGPT will show individual products and ask the user to indicate whether they like or dislike the products. If a user dislikes an item, ChatGPT asks for quick feedback on why that might be — style, price or otherwise.
You May Also Like
Based on the user’s preferences, ChatGPT shows a final list of recommendations, complete with commentary about the specifics of each item and why it matches what the user sought out. For some fashion and apparel searches, it also includes recommendations on how to coordinate the final item.
While OpenAI has long enabled users to find items on its platform, the update, powered by a version of its GPT-5 mini model that is trained on shopping-specific tasks, is meant to help with more complex inquiries. Shopping research can also search for the best deals on specific items, find lookalikes based on images submitted by the user and aid with gifting.
When sharing its recommendations, ChatGPT can draw on memory, if the user has it turned on. That means that it can draw on information from prior exchanges with the user. So, if a consumer asks ChatGPT to help find a gift for their dad, who likes golf and drinks tea every morning, the model may also remember from a previous conversation that the user said their dad enjoys watching football on Sundays and that his favorite color is red.
OpenAI researchers told reporters at a press event that the model takes organic reviews and other high-fidelity web content into account when recommending specific products for consumers and noted that, in building shopping research, they have steered the model specifically toward trustworthy sources.
Manuka Stratta, a member of OpenAI’s technical staff, told reporters at an event that ChatGPT can understand overt and covert intent from users.
“[Shopping research] will actually look at dozens and dozens of websites and sources, and it will really index on all the both implicit and explicit requirements that you’re giving it,” Stratta said.
Major retailers and brands, like Walmart, Etsy and some Shopify merchants, have started partnering with OpenAI to enable checkout directly on the ChatGPT platform through an experience the technology company calls Instant Checkout. But despite those partnerships, ChatGPT will not surface results from those retailers first.
Isa Fulford, a member of OpenAI’s technical staff, said shopping research is focused on finding a strong match for users’ inquiries.
“We don’t have any prioritization for any specific sites. It’s just what the model finds and thinks is the best option or the best deal,” Fulford told reporters.
Fulford said that today, Instant Checkout focuses on a different part of the consumer funnel than shopping research. As such, the two experiences are not currently integrated — that is to say, if a user chooses to buy an item ChatGPT surfaces in a shopping research-based query, they will still have to click out onto a brand or retailer’s own site to complete the transaction. According to an OpenAI blog post, the company does plan to integrate Instant Checkout and shopping research in the future.
Though a broad swath of brands and retailers’ items will be eligible to be shown to consumers by ChatGPT, Fulford said that the technology is trained to respect companies that block its research function from accessing their sites. That could be applicable to e-tail giant Amazon, which is actively engaged in legal action against Perplexity over its agentic shopping experience, Comet, and has actively worked to prevent third-party technology companies’ large-language models (LLMs) from accessing its product catalogue.
Amazon has been developing its own AI shopping assistant, which it calls Rufus, alongside some other AI-enabled shopping experiences. Stratta told reporters she sees Rufus and ChatGPT shopping research as “complementary” tools.
“Both platforms definitely are complementary,” Stratta said. “Maybe for Amazon, you might prioritize just a quick buy — you want to find something where it’s relatively low effort. But I think our experience really shines for the in-depth searches, where you really want to do a lot of research. Instead of just opening 30 tabs and looking through reviews yourself, this is a really great way to aggregate all the information for you.”
While some companies may not have chatbot blockers installed in the backends of their sites, researchers said the model is trained to surface high-quality results, which means that, for many queries, it may avoid budget shopping platforms like Temu.
Stratta said that, if a consumer specifically requests that their results come from discount retailers or low-cost platforms like Temu, shopping research will respect that request.
Shopping research started rolling out Monday to all logged in ChatGPT users globally and will continue to roll out throughout the week. The company said in a blog that it will make “nearly unlimited usage available to all plans through the holidays” as a way to help with gifting and inspiration.
Fulford said that OpenAI plans to continually upgrade the shopping research experience, but noted that the company has already seen high demand for such a tool.
“Already, hundreds of millions of people ask ChatGPT for product advice, and we see this especially in more spec-heavy categories like electronics and health and beauty and home…because for these kind of spec-based things, you need to compare a lot of [items] that you might not know that much about, and you want something that’s going to tell you the best thing to buy for your budget constraint preferences,” she said.