Where the products actually come from
OpenAI does not publish a full ranking algorithm, but its shopping help documentation and merchant announcements describe three pipelines feeding a shopping answer. None of them is “pay us and you appear.”
| Pipeline | What it is | Who controls it |
| Web search & crawling |
ChatGPT search uses third-party search providers — Bing is the one OpenAI has named in its docs — plus OpenAI’s own crawler, OAI-SearchBot, which reads public retail pages directly. |
You: crawlable pages, Bing indexing, on-page content. |
| Structured product metadata |
OpenAI says shopping results are built from structured metadata — pricing, descriptions, reviews — from third-party providers and merchants (TechCrunch, Apr 2025). Merchants can also apply to submit a direct product feed via OpenAI’s merchant program. |
You: schema, feed data quality, feed application. |
| Platform partnerships |
Shopify announced in September 2025 that it powers product discovery in ChatGPT with real-time pricing, inventory, images, and variants for its merchants, built on the Agentic Commerce Protocol. Etsy joined at the same launch. |
Partly you: clean catalog data flows through automatically. |
The model then does the last step itself: it interprets the shopper’s intent — budget, use case, constraints mentioned in the conversation, and (where enabled) ChatGPT’s memory of the user — and picks products it judges relevant. Since November 2025 there is also a deeper shopping research mode that builds a buyer’s-guide-style comparison before recommending.
How ranking works, as far as OpenAI has said
Per OpenAI’s help center and its Instant Checkout announcement:
- Product results are organic. OpenAI describes them as “organic and unsponsored,” ranked on relevance to the user — not ads, and not influenced by partnerships.
- Relevance to the conversation comes first. The query, stated constraints, and user context (memory, custom instructions) shape which products appear at all.
- When several merchants sell the same item, OpenAI has said it considers availability, price, quality signals, and whether the seller is the maker or primary seller of the product.
- Reviews and third-party content matter. OpenAI cites reviews and editorial sources as inputs, which means what other sites say about your product influences results you do not directly control.
What OpenAI has not published: the weighting of any of these factors, how often product data is refreshed from crawls versus feeds, or category-level differences. Anyone selling you a guaranteed “rank #1 in ChatGPT” formula is guessing. We say that as a company that sells AI-visibility tooling: the honest play is to fix the checkable inputs, then measure.
The 2026 pivot: discovery in ChatGPT, checkout on your store
Instant Checkout — buying inside ChatGPT — launched in September 2025 with Etsy sellers, with Shopify merchants following. By March 2026, CNBC and Search Engine Land reported OpenAI was moving away from the standalone in-chat checkout: shoppers were researching in ChatGPT but finishing purchases on merchant sites, so OpenAI is prioritizing discovery in ChatGPT with checkout on the retailer’s own site or app.
For a normal store this is good news. You do not need to win a checkout integration — you need ChatGPT to find, understand, and trust your product pages, because the click still lands on your storefront.
What your store can actually control
- Don’t block the crawlers. Check robots.txt for rules that block
OAI-SearchBot or ChatGPT-User (OpenAI documents its bots here). Blocking GPTBot only opts you out of training data, but blocking the search bots removes you from ChatGPT’s live reading of your site.
- Be indexable by Bing. ChatGPT search leans on third-party providers, so a store invisible to Bing is fighting with one pipeline tied off. Verify in Bing Webmaster Tools.
- Ship real structured data. Product, Offer, and Organization schema is the machine-readable layer both search engines and AI systems parse — this is the proven piece. Our free product schema generator covers the basics. An llms.txt file is an optional extra: an emerging convention that offers AI systems a curated map of your site, though no major AI crawler has publicly committed to reading it — treat it as cheap insurance, not a ranking lever.
- Fix thin product content. The model recommends what it can understand. Boilerplate two-line descriptions give it nothing to match against a specific request like “waterproof, under $100, for wide feet.”
- Keep feed data accurate. On Shopify, pricing, inventory, and variants flow to ChatGPT through the partnership — stale or contradictory data is worse than no data. Non-Shopify stores can apply for a direct feed via OpenAI’s merchant page.
- Earn third-party corroboration. Reviews and independent mentions are ranking inputs you influence slowly — but they compound.
The same inputs feed other AI shopping surfaces too: Perplexity runs its own free merchant program for sharing product data, and Google’s AI results read the same schema. Fix the layer once, and every engine benefits.