OpenAI launched Shopping Research in ChatGPT yesterday, and it's actually good. You describe what you're looking for—"find the quietest cordless vacuum for a small apartment" or "I need a gift for my four-year-old niece who loves art"—and it builds a personalized buyer's guide in a couple minutes. The interface is clean, the recommendations are solid, and it's free for everyone through the holidays.
How It Actually Works
Shopping Research uses a fine-tuned version of GPT-5 mini specialized for e-commerce tasks. Unlike regular ChatGPT, which just gives you text responses, Shopping Research opens a visual interface where you can chat and swipe through product suggestions like you're on a dating app.
You start by describing what you need. ChatGPT asks clarifying questions about budget, features, who it's for, and what matters most to you. If you have memory turned on, it factors in stuff it already knows—like if you're into gaming, it'll weight that when recommending a laptop.
Then it searches across the internet for current pricing, availability, specs, reviews, and images. As it finds options, you can mark them "not interested" or "more like this" to guide the research in real time. The model adapts based on your feedback, refining results as it goes.
After a few minutes, you get a curated buyer's guide with top products, key differences, pros and cons, and links to purchase. OpenAI says it performs especially well in detail-heavy categories like electronics, beauty, home and garden, kitchen appliances, and sports equipment.
The Amazon Problem
Here's the weird part: Amazon products barely appear in Shopping Research results. When I tested it by asking for "laptops under $1,000," ChatGPT surfaced options from Best Buy, B&H Photo, Newegg, and manufacturer sites. No Amazon links.
When I explicitly asked if it could show me Amazon listings, ChatGPT said "I can't browse or display live product listings from Amazon." If I request specific products available on Amazon, it shows alternatives from other retailers and suggests I "manually check if they're available on Amazon."
The reason? Amazon's robots.txt file restricts how AI crawlers can access its product data. OpenAI is respecting those restrictions, which means Shopping Research can't scrape Amazon's catalog the way it scrapes other retailer sites.
For users, this is annoying. Amazon has the largest product selection and often the most reviews. For OpenAI, it's strategically limiting. For Amazon, it's a defensive move to keep AI chatbots from becoming the new product discovery layer that bypasses Amazon's own search and recommendation systems.
What OpenAI Is Actually Optimizing For
Shopping Research prioritizes "high-quality, trustworthy websites" with organic content. That means review sites, Reddit discussions, and editorial content rank higher than paid marketing or product pages.
Reddit threads where real people discuss products they've actually used get weighted more heavily than sponsored reviews on a retailer's site. That's a smart trust signal, and it makes the recommendations feel more authentic than Google Shopping results or Amazon's "recommended for you" algorithm.
OpenAI explicitly says results are organic and unsponsored. Merchants can't pay to appear in Shopping Research. They can follow an allowlisting process to ensure their site is crawlable, but they can't influence rankings.
There's no checkout feature yet, though OpenAI says integration with their existing in-chat checkout tools is coming. Right now you get recommendations and links, then you leave ChatGPT to complete the purchase.
The Accuracy Question
OpenAI acknowledges Shopping Research makes mistakes. Pricing can be wrong. Availability can be outdated. Product specs might be inaccurate. They recommend verifying details on merchant sites before purchasing.
In their internal evaluations, Shopping Research hit 52% product accuracy on multi-constraint queries, compared to 37% for regular ChatGPT Search. Product accuracy means meeting all the user's requirements for attributes like price, color, material, and specs.
Fifty-two percent sounds bad until you realize how hard these queries are. "Find a waterproof Bluetooth speaker under $75 with at least 10-hour battery life in red" involves checking multiple attributes across hundreds of products. Getting it right half the time is actually decent for an AI system.
But it means you can't trust Shopping Research blindly. It's a research tool, not a purchase decision. You still need to verify before buying.
The Target Partnership
Last week, OpenAI and Target announced plans to launch Target's app inside ChatGPT. Despite that partnership, OpenAI insists Shopping Research won't favor Target in recommendations. Results are ranked by relevance to the query, not by business deals.
That's the claim. We'll see if it holds up long-term. Once OpenAI starts monetizing Shopping Research—and they will—the pressure to favor partners who pay will be enormous.
The Business Model Question
Shopping Research is free right now with "nearly unlimited usage" through the holidays. That's customer acquisition. Get people hooked on AI-powered shopping during the busiest retail season, then figure out monetization later.
The obvious revenue model is affiliate commissions. Every link you click through could include an affiliate tag, and OpenAI takes a cut of purchases. That's how every other shopping comparison tool makes money.
The other model is charging merchants for premium placement or "verified merchant" badges. Or requiring a ChatGPT Plus subscription for advanced Shopping Research features. Or all of the above.
For now, it's free and unmonetized. Enjoy it while it lasts.
How It Compares to Google Shopping
Google launched AI shopping features earlier this year, and they're... fine. Functional but not compelling. Shopping Research feels more conversational and interactive. The swipe-to-refine interface is genuinely useful, and the clarifying questions help narrow down what you actually want.
Google has the advantage of massive product catalog data and direct partnerships with retailers. OpenAI has the advantage of better natural language understanding and a more conversational interface.
The real competition isn't for who has the better AI shopping tool. It's for who captures the product discovery moment before users even get to Amazon or Google. If people start asking ChatGPT what to buy instead of searching Google or browsing Amazon, that's a fundamental shift in e-commerce behavior.
What Retailers Should Be Worried About
AI chatbots are becoming the new product discovery layer. Instead of scrolling through Amazon's search results or clicking through Google Shopping ads, users describe what they want and get curated recommendations.
That's great for consumers. It's terrible for retailers who rely on paid placement, search ranking optimization, and recommendation algorithms to drive sales. If AI decides what products to show based on "trusted reviews and organic content," all that paid advertising becomes irrelevant.
The retailers who benefit are those with strong organic reputations, detailed product information, and easy-to-crawl websites. The retailers who lose are those dependent on paid search, aggressive SEO, and algorithmic gaming.
This is the SEO apocalypse everyone's been predicting, except it's happening in product search first.
My Take
I tested Shopping Research for an hour yesterday and it's legitimately useful. Not perfect—it recommended a few products that were out of stock or miscategorized—but good enough that I'd use it for research on my next major purchase.
The Amazon gap is the biggest limitation. Until OpenAI and Amazon figure out their data access situation, Shopping Research will always be incomplete. But maybe that's a feature, not a bug. Forcing users to look beyond Amazon might surface better products from smaller retailers.
The free period through the holidays is smart marketing. People are shopping right now, and if ChatGPT becomes part of their research workflow, that's sticky behavior that persists after the holidays end.
What I'm watching is whether this changes actual purchase behavior or if it's just another research step people use before buying on Amazon anyway. If it's the latter, Shopping Research is useful but not transformative. If it's the former, we're witnessing the beginning of a major shift in how e-commerce discovery works.
My guess: it'll be somewhere in the middle. Shopping Research becomes a common first step for complex purchases where you need detailed comparisons. But quick, simple purchases still happen directly on Amazon or Google. The question is where the line falls and how much of the market that represents.