ChatGPT Product Discovery: Merchant Feed Checklist for AI Shopping Shortlists
ChatGPT processes 50 million shopping queries daily. This merchant feed checklist covers the required fields, proof attributes, sync cadence, and crawler access that determine whether your products make the AI shortlist.
ChatGPT now processes roughly 50 million shopping queries every day — about 2% of all daily prompts. (Dataslayer) Whether your products appear in those results depends on your merchant feed: the structured data file that tells ChatGPT what you sell, at what price, and with what proof. This is the checklist I use to audit feed readiness for AI product discovery.
How ChatGPT product discovery works
ChatGPT shopping is not keyword matching. It is constraint-based retrieval: a buyer asks for "waterproof hiking boots under $150 with good ankle support," and OpenAI's system matches that prompt against structured merchant data — prices, specs, reviews, availability, images. Merchants share product feeds through the Agentic Commerce Protocol (ACP), jointly maintained by OpenAI and Stripe, so ChatGPT can surface richer results with side-by-side comparisons. (OpenAI) The system achieves 64% accuracy matching products to requirements, compared to 37% for standard queries — a 73% improvement in product-intent resolution. (Dataslayer)
Why the merchant feed is the shortlist gate
The feed is your catalog's machine-readable identity. ChatGPT does not crawl your site for product data the way Google does — you push a structured file (CSV, TSV, XML, or JSON) to a secure OpenAI endpoint. (ACP Specification) Products without a feed can still appear from crawled pages, but feed-submitted products get richer cards, comparison eligibility, and eventually instant checkout. AI referral traffic to U.S. retail sites grew 393% year over year in Q1 2026. (Elogic Commerce) Your catalog should not be missing from that growth.
Required feed fields for ChatGPT eligibility
These fields determine whether a product can appear at all in ChatGPT shopping results:
| Field | What to include | Why it matters |
|---|---|---|
| Product title | Specific, attribute-rich title | ChatGPT matches buyer constraints against titles |
| Description | Features, specs, use case, materials | Evidence for constraint matching beyond the title |
| URL | Canonical product page link | Where ChatGPT sends the buyer |
| Image | High-resolution primary product image | Required for visual browsing and comparison cards |
| Price | Current price with currency | Eligibility for budget-constrained queries |
| Availability | Real-time stock status | Prevents stale recommendations and trust erosion |
OpenAI's merchant onboarding page lists these as the baseline for product card eligibility. (OpenAI Merchants)
Proof fields that win comparison cards
Required fields get you into the system. Proof fields determine your position in the comparison. When ChatGPT builds a side-by-side product shortlist, the listing with richer metadata earns more evidence surface in the comparison card:
- Reviews and rating count — ChatGPT uses social proof to differentiate competing products in constraint-heavy prompts
- Shipping speed and cost — a buyer who says "need it by Friday" eliminates products without delivery data
- Return policy — reduces perceived risk in the comparison card
- Warranty and compatibility — answers the second-order questions buyers ask after the initial shortlist
- Use-case language — "best for trail running" or "designed for home office" matches natural-language prompts directly
One in five Walmart referral clicks now comes from ChatGPT, and that share grew 15% month over month. (Dataslayer) Walmart's feed includes deep proof fields across its top SKUs. That is not a coincidence.
Feed freshness and sync cadence
Stale data is worse than missing data in ChatGPT shopping. A product shown as available at $89 that is actually out of stock at $109 creates shopper distrust, and ChatGPT will deprioritize unreliable feeds over time. The ACP specification accepts feed refreshes as frequently as every 15 minutes. (ACP Specification)
My recommendation: sync daily at minimum for catalog-wide updates, and push real-time updates for price changes, stock-outs, and promotions on your top 100 revenue products. If your feed management platform (Feedonomics, DataFeedWatch, Lengow) supports scheduled ACP pushes, configure them.
Crawler access and discovery permissions
Even with a feed, ChatGPT's web crawler OAI-SearchBot needs access to your product pages, category pages, and supporting content. OpenAI's guidance says merchants should avoid blocking OAI-SearchBot if they want products discoverable in ChatGPT search. (OpenAI)
Amazon blocked OAI-SearchBot and saw referral traffic drop 18% month over month to sub-3%. (Dataslayer) Meanwhile, Target, Etsy, and eBay — which kept crawlers open — saw rising ChatGPT referral share. The feed is the product identity; the crawler is the context discovery layer. Block one and the other weakens.
The weekly feed health audit
I run this five-point check every Monday:
- Field completeness — Export top 100 revenue SKUs. Flag every missing attribute a buyer would use in a natural-language prompt.
- Price and stock accuracy — Compare feed values against live PDP. Any mismatch is a trust defect that compounds across queries.
- Proof field coverage — Count SKUs missing reviews, shipping data, return policy, or use-case language. Target: 80% coverage within 30 days.
- Freshness lag — Measure the gap between feed push and last inventory change. Anything over 24 hours on a top-100 SKU is exposure.
- Crawler health — Check server logs for OAI-SearchBot activity. If crawl rate dropped, investigate robots.txt or server-side blocks.
This is Machine Relations work, not SEO. You are not optimizing a page for a ranking algorithm. You are making your catalog legible enough for an AI system to retrieve, compare, and recommend. The feed is the structured evidence layer; the product pages are the entity chain that backs it up.
Competitive position in AI shopping shortlists
Here is the dynamic most operators miss: ChatGPT shopping is zero-sum at the comparison level. When a buyer asks for "best wireless earbuds under $100," ChatGPT builds a shortlist of three to five products. A competitor with a more complete feed wins the comparison card even if your product is objectively stronger.
Merchants implementing multiple agentic commerce protocols — ACP, Google UCP, Apple Commerce — see approximately 40% more agentic traffic than single-protocol adopters. AI platforms are projected to drive $20.9 billion in retail spending in 2026, nearly quadrupling 2025. (Elogic Commerce) Over 1 million Shopify merchants are coming to ChatGPT Instant Checkout. (Dataslayer)
The window to establish feed completeness ahead of that wave is now. Use the AuthorityTech visibility audit to test whether your brand already appears where buyers are asking.
FAQ
What is ChatGPT product discovery?
ChatGPT product discovery is OpenAI's AI shopping experience that surfaces products through conversational, visual, and comparison-style results. It uses merchant feed data, Agentic Commerce Protocol integrations, and web-crawled content to match products to buyer constraints like budget, features, and intended use.
Do I need a product feed for ChatGPT shopping?
Not strictly. ChatGPT can discover products through OAI-SearchBot web crawling. But a structured ACP feed gives you richer product cards, comparison eligibility, and faster data updates. Without a feed, you depend entirely on what the crawler finds on your pages.
How often should I update my ChatGPT product feed?
The ACP specification accepts refreshes every 15 minutes. At minimum, push daily full-catalog updates and near-real-time updates for price changes and stock-outs on your highest-revenue products.
What happens if I block OAI-SearchBot?
Your products lose context-layer discoverability in ChatGPT search. Amazon's referral traffic from ChatGPT dropped 18% month over month after blocking OAI-SearchBot, while Target, Etsy, and eBay gained share by keeping crawlers open. (Dataslayer)
Is this SEO or Machine Relations?
This is Machine Relations work. SEO optimizes pages for ranking algorithms. Machine Relations makes structured evidence, entity data, and trust signals clear enough for AI discovery systems to retrieve and recommend. The merchant feed is the structured evidence layer for AI product discovery.
How do I check if my products appear in ChatGPT?
Ask ChatGPT product-comparison prompts using the constraints your real buyers would use. Test your top 10 SKUs against named competitors. If your products do not appear, start with the feed completeness and crawler access sections of this checklist.