The OpenAI Product Feed: Getting Your Products Into ChatGPT Shopping

Why the Feed Matters More Than Checkout Now

When OpenAI launched Instant Checkout inside ChatGPT in September 2025, the pitch was buying products without leaving the chat. That pitch has changed. In early 2026, OpenAI scaled back native in-chat checkout and refocused ChatGPT Shopping on product discovery, sending shoppers to merchant-owned sites and apps to complete the purchase.

For store owners, this is good news. You don't have to re-plumb your checkout to participate. What you do need is a clean, structured product feed that OpenAI can ingest and index — because that feed is what determines whether your products show up when someone asks ChatGPT for a recommendation. Discovery happens in the model; the sale happens on your site.

If you've been treating ChatGPT visibility as purely a content-and-schema problem, the feed is the missing piece. It's a direct, machine-readable pipeline into the recommendation surface.

What the OpenAI Product Feed Is

The feed is a flat-file catalog you submit to OpenAI following its published product feed specification. OpenAI ingests and indexes it, then draws on it when generating shopping answers. Think of it as the ChatGPT equivalent of a Google Merchant Center feed — but built for a conversational engine that surfaces one or two products, not a grid of twenty.

Two things make it different from traditional shopping feeds:

It's a push model. Rather than pointing a crawler at your site, you deliver structured files to OpenAI. Approved merchants can refresh frequently to keep pricing and availability current.
You control eligibility per product. Fields like is_eligible_search let you decide which items appear in ChatGPT shopping results, giving you catalog-level control over what gets surfaced.

The Onboarding Paths

There are a few ways in, and the right one depends on your platform:

Already on Shopify or Etsy? Your catalog is integrated through the platform, so there's little or no separate setup — the heavy lifting is handled by the integration. Confirm the ChatGPT/commerce channel is enabled in your admin.
On another platform or headless? Apply through OpenAI's merchant portal. OpenAI verifies your business and approves you for feed ingestion before your products can appear.
Feed management tooling. Many merchants generate and maintain the feed through a feed-management or PIM tool that outputs the required schema, then hand delivery to OpenAI.

Whichever path you take, the feed schema underneath is the same — so the quality of your product data is what ultimately decides how well you show up.

The Required Fields

OpenAI's spec defines a core set of attributes every item needs. Get these right before worrying about anything optional:

item_id — your stable product identifier (max 100 characters). Don't recycle IDs across different products.
title — product name, up to 150 characters. Lead with what the product is, not marketing adjectives.
description — full plain-text details, up to 5,000 characters. This is where specificity pays off.
url — the product detail page, which must resolve with an HTTP 200.
brand — the product's brand (max 70 characters).
image_url — main product image (JPEG or PNG).
price — regular price with an ISO 4217 currency code.
availability — one of in_stock, out_of_stock, pre_order, backorder, or unknown.
seller_name and seller_url — who's selling and where.
target_countries and store_country — ISO 3166-1 alpha-2 country codes.

Alongside those, discovery flags such as is_eligible_search control whether an item is surfaced in ChatGPT shopping results. Setting these deliberately keeps discontinued or out-of-region SKUs out of your visible catalog.

The Optional Fields That Actually Move the Needle

Meeting the required minimum gets you indexed. The optional fields are what make you recommendable:

gtin / upc — universal identifiers help the model match your item to the same product mentioned elsewhere, reinforcing that your listing is legitimate.
sale_price — surfaces active discounts, which matters when a shopper asks for the best deal.
variant_dict — size, color, and other options, so the model can answer "do they have it in medium?" without guessing.
star_rating / review_count — social proof the model weighs when choosing between comparable products. This mirrors how reviews and ratings shape AI recommendations across every engine.
shipping — fulfillment and delivery details, part of the trust signals AI engines read when deciding which store to name.
warning — compliance disclaimers where products require them.

The pattern here is the same one that governs all AI search visibility: structured, specific, verifiable data beats vague marketing copy. A feed that fills in GTINs, variants, ratings, and precise availability gives the model concrete reasons to choose you.

Keep the Feed Fresh

A stale feed is worse than no feed. If ChatGPT recommends a product that's out of stock or priced wrong, you burn trust with both the shopper and the engine. Because the feed supports a push model, treat it like live infrastructure:

Sync price and availability on a tight cadence, especially during promotions.
Remove or flag discontinued SKUs promptly rather than letting them linger.
Make sure every url still resolves with a 200 — dead product pages undermine the whole feed.

Feeds that update frequently and stay accurate are the ones that earn durable placement.

Where the Feed Fits in a Broader Strategy

The product feed is powerful, but it isn't a standalone solution. ChatGPT cross-references the feed against everything else it knows about your brand — your on-page content, your structured data, and third-party mentions. A feed with immaculate data still underperforms if the model can't corroborate your brand as a real, trusted entity elsewhere on the web.

Think of it as three reinforcing layers:

The feed delivers clean, structured product facts directly to OpenAI.
On-page schema and content let the model verify those facts when it browses your site.
Off-site signals — reviews, mentions, and citations — confirm your brand is legitimate and worth recommending.

Stores that align all three are the ones ChatGPT names. If you're not sure where your catalog stands, an AI visibility audit will show which products are eligible, where your feed data is thin, and what's blocking you from being surfaced.

The Takeaway

ChatGPT Shopping has settled into a clear shape: discover in the AI, buy on your site. The product feed is how you get into the discovery layer. Submit a complete, accurate feed that hits every required field, invest in the optional fields that build trust, and keep it fresh. Then make sure your product data and schema back it up when the model comes to verify. That's the path to being the store ChatGPT recommends — and the traffic lands on your checkout, where it converts best anyway.