What Is ChatGPT Shopping — and How Does It Choose Products?

ChatGPT isn’t just answering questions about products anymore. It’s recommending them — with names, prices, images, and direct purchase links. When a shopper asks “what’s a good travel stroller under $300?” ChatGPT returns a curated set of specific products, not a list of links to browse.

This changes the game for merchants. Your product is either in that recommendation or it isn’t. There’s no page two.

How ChatGPT discovers products

ChatGPT Shopping doesn’t crawl the internet in real time for every query. It pulls from multiple data sources, each with different mechanics.

Google Merchant Center feeds. This is the primary pipeline. ChatGPT’s product recommendations lean heavily on the same product data that powers Google Shopping. If your Merchant Center feed is incomplete, stale, or poorly structured, ChatGPT inherits those problems. A product with a vague title like “Blue Shoe” in your feed will lose to a competitor listing “Nike Air Zoom Pegasus 41 Women’s Running Shoe, Wide, Size 9, Thunder Blue.”

On-page JSON-LD markup. ChatGPT’s web browsing capability reads structured data from product pages. JSON-LD Product markup tells ChatGPT the price, availability, brand, identifiers, and reviews in a format it can parse without guessing. Pages without JSON-LD force the model to extract data from HTML — which is slower, less reliable, and often wrong.

Third-party data partnerships. OpenAI has data-sharing agreements with shopping platforms and aggregators. The exact partners aren’t fully public, but the pattern is clear: if your products appear in major shopping databases and comparison engines, they’re more likely to surface in ChatGPT results.

Web crawling and indexing. ChatGPT can browse the web, and when it does, it reads what any crawler would read: your page content, meta tags, structured data, and linked resources. But browsing is slow and expensive compared to querying a feed. Products available through structured feeds get priority.

What data ChatGPT uses to make recommendations

When ChatGPT decides which products to recommend, it evaluates several signals.

Product-query match. Does this product actually answer what the shopper asked? ChatGPT matches product attributes against the query. A shopper asking for “waterproof hiking boots for wide feet” needs a product with explicit waterproof rating, hiking classification, and wide-width availability in its data. If those attributes are buried in a paragraph of marketing copy instead of structured fields, the match is weaker.

Price and availability. ChatGPT filters for products that are currently in stock and within the shopper’s stated budget. Stale inventory data — showing “in stock” when you’re actually sold out — burns trust fast. ChatGPT learns which merchants have reliable availability data and which don’t.

Reviews and ratings. Products with AggregateRating markup in their JSON-LD get a trust boost. ChatGPT can surface star ratings and review counts directly in its recommendations. A product with 4.6 stars from 2,300 reviews is more likely to be recommended than one with no reviews at all.

Brand verification. ChatGPT is more confident recommending products when it can verify the brand across multiple sources. If your brand name is consistent across your website, Merchant Center feed, and review platforms, the model treats it as a real, established brand. Inconsistencies — “Acme” on your site, “Acme Co.” in your feed, “ACME Athletics” on Amazon — erode that confidence.

Attribute specificity. Vague products lose. A listing that says “comfortable running shoe” provides nothing for ChatGPT to work with when a shopper asks for “lightweight neutral-cushion running shoe under 8 ounces for marathon training.” The merchant whose listing includes weight, cushion type, drop height, and intended use wins.

What merchants can do right now

Audit your Merchant Center feed. This is job one. Check every product for complete titles (brand + product type + key attributes), accurate pricing, real-time inventory, and valid GTINs. A feed audit reveals most of the gaps that make you invisible to ChatGPT.

Add JSON-LD to every product page. At minimum: Product type with name, brand, SKU, GTIN, image, description, and a complete Offer block with price, currency, availability, and condition. Add AggregateRating if you have reviews.

Write attribute-dense descriptions. Replace “our best-selling shoe” with specific, measurable attributes. Weight, dimensions, materials, compatibility, use cases. These are the fields ChatGPT matches against natural-language shopping queries.

Keep inventory accurate. If ChatGPT recommends your product and the shopper clicks through to find it out of stock, that’s a negative signal. Real-time inventory sync between your store and your feeds isn’t optional anymore.

Maintain brand consistency. Use the exact same brand name everywhere — your site, your feeds, your social profiles, your marketplace listings. This is how ChatGPT verifies you’re a real brand worth citing by name.

The shift that matters

ChatGPT Shopping turns product data into a direct sales channel. Not a marketing channel, not a brand awareness play — a channel where a shopper goes from question to purchase recommendation in seconds, and your product data is the only thing standing between you and that recommendation.

Merchants who treat their structured data as the storefront it now is will show up. Everyone else will wonder where their traffic went.