How AI shopping assistants choose which products to recommend

When a shopper asks an AI shopping assistant “what’s the best rain jacket under $200?”, it doesn’t return ten blue links. It returns a few products and a reason for each. Which products make that list, the AI shelf, is quickly becoming as important as search rankings. Here’s what the platforms themselves say about how it works.

Why it matters now

Shopping through AI assistants is still small compared with search, but it’s growing quickly. Adobe measured traffic from generative AI tools to US retail sites up 693% year over year in the 2025 holiday season. Meanwhile, ChatGPT, Google, Perplexity and Microsoft Copilot have all added shopping features, and several now let shoppers buy without leaving the conversation, or hand them straight to the store’s checkout.

What the platforms say

ChatGPT

OpenAI’s help center says product results “are selected independently by ChatGPT and are not ads”, with ads kept separate. A product appears “when ChatGPT perceives it to be relevant” to the question and context. ChatGPT draws on structured product data from merchants and other providers, such as price and description, and on other third-party content. When several stores sell the same product, OpenAI says merchants “are ranked based on factors like availability, price, quality, and whether they are the maker or primary seller of that item.”

In 2026, OpenAI moved away from a standalone checkout inside ChatGPT: its merchant page says purchases are now completed on merchants’ own websites or apps. Its shopping research feature reads product pages directly, cites its sources, and warns that it can make mistakes about price and availability.

Google

Google’s AI shopping features draw on its Shopping Graph, which Google says contains more than 50 billion product listings, 2 billion of them updated every hour. Google has added agentic checkout for eligible US merchants, asking the shopper’s permission before buying. Google hasn’t published how its AI answers rank products.

Perplexity and Microsoft Copilot

Perplexity offers shopping with checkout through PayPal, with the retailer remaining the seller. Microsoft’s Copilot Checkout, announced in January 2026, lets shoppers buy inside Copilot while the retailer keeps the sale and the customer relationship. Neither has said in detail how products are ranked.

If you sell on Shopify

Shopify’s agentic storefronts make eligible stores’ products available in AI channels including ChatGPT, Google’s AI Mode and Gemini, Microsoft Copilot and Meta, and Shopify says the feature is on by default for eligible stores. Product data flows through Shopify Catalog, which keeps prices and stock up to date. Two things worth knowing:

• You can manage which AI channels receive your products under Sales channels → Agentic in Shopify admin.
• Shopify notes that blocking AI crawlers in robots.txt doesn’t stop Shopify Catalog from sending your product data.

Being in the catalog makes you eligible to be recommended. It doesn’t guarantee you will be.

What you can influence

None of the platforms sell places in their organic shopping answers, and none publish a full ranking formula. But their guidance points in the same direction:

• Complete, accurate product data: titles, specifications, variants, images, price and availability, in your store and your product feed.
• Clear, competitive prices and policies: OpenAI names price and availability among its ranking factors, and assistants often quote shipping and returns.
• Genuine reviews, on your site and on independent sites.
• Presence on the sources assistants cite: review sites, buying guides and trusted retailers. Look at the citations in answers for your category.
• Pages AI agents can read: if agents can’t load your product pages or understand your prices, they can’t recommend you accurately. Our sister product Ghost Agent Labs tests this.

Measure it, repeatedly

Answers vary between runs, users and weeks, and a competitor’s price cut or a new review can change them. Treat your place on the AI shelf as something to measure over time, next to what your competitors are doing. See how to measure your share of the AI shelf.