How to Add Your Products to ChatGPT Shopping (2026)

1. Audit your catalog for AI agent readiness

Before you submit anything, identify the gaps. AI agents evaluate structured product data: complete attributes (size, color, material, fit, dimensions), accurate pricing, real-time inventory, clear category taxonomy, and natural-language descriptions. Most catalogs score 30-60% on attribute completeness; run an AI Readiness check to see your starting point.

2. Format your catalog to OpenAI's ChatGPT product feed spec

OpenAI's feed format is JSONL with required and optional fields. Required: product ID, title, description, price, availability, image URL, category. Optional but high-leverage: structured attributes (size, color, material, dimensions), variants, ratings, GTIN, brand, condition. Shopify generates this automatically through Agentic Storefronts (April 2026 expansion to all 5.6M Shopify stores). Other platforms typically need either a Stripe ACP integration, a custom feed pipeline, or an infrastructure partner that handles product feed management across multiple AI surfaces simultaneously. Full spec at the ChatGPT product feed format glossary entry.

3. Enable ACP compliance

ChatGPT Shopping runs on the Agentic Commerce Protocol (ACP), an open standard by Stripe and OpenAI. ACP defines the handshake between AI agents and your catalog: how products are discovered, how attributes are exchanged, and how the user gets redirected to your site for checkout. Without ACP-compliant endpoints, your products are invisible to ChatGPT regardless of how rich your catalog is internally.

4. Submit and verify

Submit your feed through OpenAI's commerce platform or your integration provider. Initial indexing takes 24-72 hours. Verify by searching for your products directly inside ChatGPT (e.g. "looking for [your product category] under $X") and confirming the click-through flow from the product card to your merchant page. If products do not appear within 72 hours, the issue is almost always catalog data quality, not submission.

5. Monitor and optimize

Going live is the start, not the end. Track your AI visibility: which queries surface your products, what position they appear in, and which competitors take share. Catalog changes that move the needle in ChatGPT often look different from what moves the needle in Google. Real-time inventory and structured attribute completeness drive ranking; bidding does not.

The most common reasons products do not appear in ChatGPT Shopping after enrollment:

1. Incomplete attribute coverage. AI agents filter by structured fields (size, color, material, fit). Products missing key attributes for their category drop out of structured queries even when the description is rich.
2. Generic or marketing-rewritten product titles. "Effortless Summer Statement" is a pretty title and a useless one for AI agents matching "white linen midi dress." Move marketing copy into descriptions; keep titles structured and product-accurate.
3. Stale inventory or pricing data. Google's Shopping Graph refreshes 2 billion listings per hour. ChatGPT prefers live data. Nightly feed cadence loses you intra-day stock changes.
4. Inconsistent category taxonomy. Mapping the same product across multiple Google product taxonomy categories or using non-standard category strings drops your ranking signal across the agent's session.

Shopify auto-enrollment got 5.6M stores onto ChatGPT's shelf overnight. Catalog quality is what determines whether products on that shelf actually appear in recommendations. Full breakdown of the four reasons.

The trade-offs: Shopify is fastest if you are already on Shopify but the showup-quality bottleneck is your catalog. Stripe ACP is the standards-based path but ACP-only - you still need a separate UCP integration for Google AI Mode. OpenAI direct is enterprise-only and slow. Infrastructure providers handle protocol distribution + catalog optimization in parallel.

1. Nightly feed cadence

The Shopping Graph refresh rate (~2 billion listings per hour) is the floor AI agents expect. A nightly feed leaves you wrong all day. Move to webhook-based or short-interval (15-minute) pulls.

2. Marketing copy as product titles

"Build for the trail" is invisible to AI agents matching "trail running shoe with rock plate." Keep titles structured and product-accurate; move marketing into descriptions.

3. Missing or wrong GTINs

GTIN is one of Google's strongest matching signals. Wrong GTIN, missing GTIN, or made-up GTIN drops you out of competitive product clusters. For private-label or handmade products without GTIN, use the explicit identifier_exists: false flag rather than blank.

4. Single-channel optimization

ChatGPT runs on ACP. Google AI Mode runs on UCP. Microsoft Copilot adopted UCP in April 2026. Optimizing only for one leaves the other half of AI shopping traffic on the table.

5. Treating it like Google Shopping ads

Bidding more does not push you up in AI agent recommendations. Structured attribute completeness, real-time freshness, and natural-language descriptions are the levers. Higher ad spend without better data does not move the needle.