Back to all
Ask ChatGPT for a waterproof running jacket under $150 and it returns a handful of specific products, not a page of links. That answer is assembled from product data: pages the assistant has read across the web, and feeds merchants submit directly to OpenAI. AI shopping runs on that data, and one of those two sources sits entirely within your control.
The short of it
- ChatGPT builds product recommendations from two sources: what it reads across the open web and what merchants submit directly through OpenAI’s product feed. The feed is the source you control.
- Ranking is organic. OpenAI weighs availability, price, quality, and whether you’re the maker or primary seller of the item, and paid placements can’t alter the recommendation.
- Required feed fields get your price and availability displayed correctly. Recommended fields like reviews, rich media, and performance signals do the ranking work, and most brands skip them.
- Discoverability and checkout are separate. A product can appear in ChatGPT’s answers without in-chat purchasing, so there’s no reason to wait for Instant Checkout to reach your category.
What is AI shopping?
AI shopping is product discovery and purchasing that happens inside an AI assistant. A shopper describes what they need in natural language, and the assistant interprets the request, compares options, and recommends a short list of specific products, in some cases with checkout built into the conversation.
The behavior is already mainstream. The new discovery engine found that 70% of shoppers used AI to shop in the past three months, and one in five bought from a brand they had never heard of because an assistant recommended it. The short list itself, the agentic shelf, rebuilds for every question and holds no reserved spots.
How does ChatGPT decide which products to recommend?
ChatGPT is the useful place to start because OpenAI publishes how the system works and accepts product data straight from merchants.
The assistant draws product information from two paths. The first is what it reads: product pages, marketplace listings, retailer sites, reviews, and third-party data providers. The second is what you send: a structured product feed submitted directly to OpenAI. Where a feed exists, OpenAI treats it as the source of truth for pricing and availability rather than reconstructing those details from crawls.
OpenAI’s shopping documentation lists availability, price, quality, and whether the seller is the maker or primary seller of the item among its ranking factors, with weighting that shifts to match the request. Mention a $30 budget and price counts for more; leave it out and other attributes carry the decision. Sponsored placements run in ChatGPT as of early 2026, but they sit on a separate system and cannot alter the organic recommendation.
ChatGPT also rewrites product titles and descriptions into simplified versions before showing them, so your copywriting rarely survives the trip. The structured attributes underneath it do.
The product feed is the part you control
OpenAI’s product feed is a structured file covering identifiers, titles, descriptions, pricing, inventory, media, and fulfillment details, pushed to OpenAI on a schedule you set. Updates are accepted as often as every 15 minutes, which is the difference between an assistant quoting this morning’s price and last week’s.
The spec splits fields into two tiers with different jobs. Required fields make sure price and availability display correctly. Recommended fields, including rich media, review data, and performance signals, improve ranking, relevance, and shopper trust. Treating the recommended tier as optional is the most common way to be technically present and practically invisible.
The feed also separates discoverability from checkout. One flag controls whether a product is searchable in ChatGPT; another controls whether it can be purchased inside the conversation. Instant Checkout is still rolling out to selected merchants and regions, and none of that rollout gates whether your products appear in answers today. Merchants apply through OpenAI’s merchant program to get a feed validated and ingested.
How to get your products recommended by ChatGPT
- Submit the feed. Apply through OpenAI’s merchant program, validate a sample file, then push regular snapshots. Until a feed exists, ChatGPT reconstructs your products from whatever it finds on the open web.
- Fill the recommended fields, not just the required ones. GTINs, variant data, rich media, review counts and ratings, and performance signals all feed the ranking. Write attribute values in the language shoppers use when they ask.
- Keep pricing and inventory current. The spec accepts refreshes as often as every 15 minutes, so there’s no structural reason for an assistant to quote yesterday’s price. When it happens anyway, the trust cost lands on the brand.
- Stay consistent everywhere else ChatGPT reads. The feed covers what you send OpenAI. Marketplace listings, retail partner pages, and your own product pages still inform answers, and conflicting prices or specs across channels read as unreliability. Keep OAI-SearchBot, OpenAI’s search crawler, unblocked so it can reach your pages at all.
- Ask ChatGPT what it recommends. Run your shoppers’ top questions monthly and record which products appear, using the answer-share method from our commerce GEO guide.
AI shopping tools beyond ChatGPT
In addition to ChatGPT, shoppers already use AI shopping tools like Google’s Gemini and AI Mode, Microsoft Copilot, Perplexity, and Amazon’s Rufus, and each builds recommendations from structured product data rather than marketing copy. Google’s path runs through the Merchant Center feed most brands already maintain. Amazon’s runs through your listing data.
That overlap is the practical good news: feed work done for ChatGPT strengthens your position on every other assistant at the same time.
Where Rithum fits
Everything above is data work, and it’s the data work Rithum already does. Rithum enriches product attributes, keeps pricing and availability in sync across 900+ retail destinations, and feeds that data directly to ChatGPT and to Google, including Gemini, so assistants work from your data rather than reconstructing it from crawls. RithumIQ flags weak attributes and proposes each fix for your team to approve.
If ChatGPT is describing your products from a crawl today, the fix starts with the feed. See how Rithum builds product feeds ready for every channel, including AI.
FAQ
What is AI shopping?
AI shopping is product discovery and purchasing that happens inside an AI assistant such as ChatGPT, Gemini, or Copilot. The shopper asks in natural language and the assistant recommends specific products, drawing on product data from the web and from feeds merchants submit directly.
How does ChatGPT choose which products to recommend?
ChatGPT matches product data against the shopper’s request and ranks results on factors OpenAI lists as availability, price, quality, and whether the seller is the maker or primary seller of the item. Results are organic, and sponsored placements cannot alter them.
How do you get your products into ChatGPT?
Apply through OpenAI’s merchant program at chatgpt.com/merchants, submit a product feed that passes validation, and keep it refreshed. Without a feed, ChatGPT builds its picture of your products from what it reads on the open web.
Do products need Instant Checkout to appear in ChatGPT?
No. Search visibility and in-chat checkout are separate settings in the product feed, so a product can be recommended in ChatGPT answers without being purchasable inside the conversation.
What are AI shopping tools?
AI shopping tools are assistants and agents that recommend or purchase products for shoppers, including ChatGPT, Google’s Gemini and AI Mode, Microsoft Copilot, Perplexity, and Amazon’s Rufus. All of them build recommendations from structured product data.
