How Does ChatGPT Rank Shopping Results?

What is documented and what is inferred?

What factors affect AI product rankings: conversion rate, reviews, inventory turnover or data freshness?

Of the four, reviews and data freshness have evidence behind them. Conversion rate and inventory turnover do not.

• Reviews. OpenAI's product feed has optional review_count and star_rating fields, plus separate store rating fields. In Profound's analysis of about 812,000 product cards from late June 2026, rank-one cards had a median of 787 reviews, against 352 for cards ranked fourth or lower.
• Freshness. Availability is a required feed field and one of the factors OpenAI names for ranking merchants. OpenAI also warns that ChatGPT can get prices and availability wrong. Stale data is a practical risk even if it is not a ranking factor in its own right.
• Conversion rate. OpenAI does not mention it, and ChatGPT cannot see your checkout data unless you share it. Treat claims that it ranks by conversion rate as speculation.
• Inventory turnover. No official or independent source supports it.

Relevance comes first. Profound observed that ChatGPT turns a shopping prompt into product sub-queries, much like the fan-out it uses for text answers. Products that match those sub-queries by name, attributes and category become candidates; the signals above then sort them.

A caution on reviews: grow them honestly by asking every customer. The FTC's rule on consumer reviews bans incentives conditioned on positive sentiment, and suppressing negative reviews.

Do open-ended prompts trigger ChatGPT Shopping more than brand-name prompts?

The trigger is commercial intent, not whether a brand is named. Profound describes shopping mode switching on when a prompt passes a commercial-intent threshold. A prompt such as "best trail running shoes under $150" and one such as "Hoka Speedgoat 6 price" both show buying intent. "Is Hoka a good company?" does not.

What differs is the result. Open-ended prompts produce a comparison across brands, where your product competes on relevance and reviews. Brand or model prompts usually show that product, and the competition moves to the offer list of merchants selling it. AEOsome's study of 8,520 shopping queries found product carousels appeared for 59.5% of them. We know of no published study comparing trigger rates for open-ended and brand-name prompts, and we have not measured it ourselves.

What makes one seller the primary seller for a product in ChatGPT?

OpenAI names "whether a merchant is the maker or primary seller" as a factor when ranking merchants for the same product, but it does not define primary seller. A reasonable reading is the brand's own store, or the retailer most closely identified with the product, such as an exclusive stockist.

In practice, the offer list matters. Profound found that when a product card has offers, its price and merchant details come from the rank-one offer. To be that offer:

• Sell the product under its exact name and identifiers, so ChatGPT can match your listing to the product.
• Keep availability and price accurate, since both are named factors.
• If you are the maker, say so clearly on the product page and in the feed's brand and seller fields.
• Offer clear shipping and returns data, which the feed spec supports; omitted shipping price is treated as unknown, not free.

Does ChatGPT Atlas cite product pages, category pages, or reviews for shopping recommendations?

Atlas was OpenAI's browser with ChatGPT built in. OpenAI retired it in August 2026 and folded its browser agent capabilities into ChatGPT. Studies of ChatGPT's shopping answers show that citations go mostly to retailers. AEOsome found retail sites made up 89% to 99% of citations across buying stages, with brand sites barely registering. Omniscient Digital's analysis of branded prompts found reviews and social proof made up 57% of citations. So the answer is: retailer product and category pages, and review content, far more than brand pages.

The agent that started in Atlas, now part of ChatGPT's built-in browser, can open pages and act on them. OpenAI's publisher guidance says accessibility helps that agent, since it uses ARIA labels and roles to understand page structure. Product pages with clear prices, variants and buttons in accessible HTML are easier for it to use.

How does ChatGPT Shopping Research choose which products to recommend?

Shopping research, launched in November 2025 for logged-in users, works differently from a quick answer. It asks clarifying questions about budget and preferences, then spends a few minutes building a buyer's guide. OpenAI says it pulls price, availability, reviews, specifications and images from high-quality sources across the web. Along the way, shoppers can flag options as "Not interested" or ask for "More like this", which narrows the guide.

It runs on a version of GPT-5 mini trained for shopping. OpenAI reported 52% product accuracy on multi-constraint queries in its internal tests, against 37% for ChatGPT Search. It works best in electronics, beauty, home and garden, kitchen and appliances, and sports and outdoors. Results are organic, and merchants who want to appear can follow an allowlisting process with OpenAI. To be chosen, a product needs specifications and reviews the research can read, on pages it can reach.

How do GTINs and pricing affect product placement in ChatGPT carousels?

• GTINs match products. OpenAI's feed spec treats gtin as optional, accepts 8, 12, 13 or 14 digits, and says not to invent one. A correct GTIN helps ChatGPT recognise that your listing is the same product other merchants sell, which is what puts you in its offer list.
• Google Shopping overlap. Semrush found that across 100 runs, ChatGPT's top product was in Google Shopping's top three 75% of the time, and Search Engine Land reported that 83% of carousel products came from Google Shopping data. Accurate Merchant Center feeds, with GTINs, feed both.
• Price. Price is a named merchant factor. Profound found promotional prices only slightly more common on top cards (1.19% against 1.05%), so discounting is not a shortcut.

What Profound found mattered more were machine-assigned product tags, such as labels for particular features, which appeared on 2.39% of top cards against 0.98% of low-ranked ones. ChatGPT assigns them from product data, so complete attributes help. Feed problems are covered in why products never appear in ChatGPT despite an accepted feed.

What should an online store do?

1. Submit a product feed to OpenAI and keep your Google Merchant Center feed accurate, with GTINs, brand, availability and price.
2. Complete product attributes buyers filter by, such as size, material and compatibility.
3. Keep reviews growing honestly, and pass review counts and ratings in the feed where allowed.
4. Make product pages crawlable and accessible, with prices and stock in the HTML.
5. Check prompts buyers use, open-ended and brand-specific, and note which merchants appear.

Broader context is in how ChatGPT is turning into a shopping platform, and competitive gaps in why ChatGPT recommends competitors instead of your online store.

What might this look like in practice?

Consider an illustrative cookware brand, Ironleaf, whose cast-iron skillet appeared in ChatGPT carousels but with a large retailer as the rank-one offer. Its own store listed the skillet without a GTIN and under a slightly different name. Ironleaf added the GTIN, matched the product name to its packaging, stated "made by Ironleaf" on the page, and submitted its feed with review counts. Within weeks, its store began appearing first in the offer list for brand-name prompts. For open-ended skillet questions, it still ranked behind two rivals with several times more reviews.

What do people ask about ChatGPT shopping rankings?

Can I pay to rank higher in ChatGPT shopping results?

No. OpenAI says product results are not ads and are chosen independently. Ads appear separately.

Does ChatGPT use Google Shopping data?

Studies found strong overlap with Google Shopping results, so accurate Merchant Center data helps.

Do I need a GTIN for ChatGPT?

It is optional in OpenAI's feed, but a correct GTIN helps match your listing to the product.

Does ChatGPT rank products by conversion rate?

There is no public evidence that it does.