What Shopify's ACP Product Feed Actually Sends to ChatGPT, And the Fields Most Stores Have Wrong

By Steve Merrill | April 18, 2026

When Shopify activated Agentic Storefronts on March 24, 2026, it started sending your product data to ChatGPT every day through OpenAI's Agentic Commerce Protocol feed. That's happening whether you've optimized for it or not.

The feed is automatic. The quality isn't.

What Shopify sends is only as good as what you've put into your product catalog. And most stores have the same five problems in their data, problems that reduce recommendation rates and cost real revenue in AI shopping.

What Does the ACP Feed Actually Contain?

Shopify's ACP feed includes: product title, description, price (and compare-at price if set), availability status, GTIN/barcode, product images (primary and additional), product URL, variant data (size, color, material), and inventory level.

This data goes to a daily endpoint that ChatGPT's product system ingests. When a user asks ChatGPT a shopping question, the system matches their query against this feed data to identify and rank candidates. According to Opascope's ACP protocol analysis, the feed updates daily, meaning stale data in Shopify directly translates to bad recommendations in ChatGPT within 24 hours.

Which Fields Are Most Stores Getting Wrong?

I've run this analysis across dozens of Shopify stores. The same problems come up every time.

Missing GTINs. Products without a barcode or GTIN (Global Trade Item Number) are harder for ChatGPT to cross-reference against other data sources. That cross-referencing is how the system validates that your product data is accurate. No GTIN means lower confidence, which means lower recommendation rate. In Shopify, GTINs go in the Barcode field under each product variant.

Descriptions that are too short or too generic. The ACP feed uses your product descriptions to match against natural language queries. A description like "High-quality leather wallet. Available in brown and black." tells ChatGPT almost nothing about who this product is for, what makes it good, or what queries it should match. A description that answers real buyer questions, "slim profile fits 6-8 cards without bulge, full-grain leather that develops a patina over time, RFID-blocking liner included", matches dozens of specific buyer queries.

Price inconsistency. If your ACP feed shows $49 but your storefront shows $39 (after a discount code or flash sale), ChatGPT can detect the discrepancy through price validation. Stores with frequent price mismatches between feed and live site get flagged as unreliable and deprioritized. Keep your Shopify prices current, don't hide deals behind discount codes that aren't reflected in the feed.

Single product images. Products in the feed with only one image underperform in ChatGPT's carousel compared to products with 3+ images. The system doesn't show all your images in the carousel, but the presence of multiple images signals a more complete and trustworthy product listing.

Vague product titles that omit key attributes. "Women's Jacket - Blue" could match hundreds of different buyer queries or none of them. "Women's Insulated Puffer Jacket, Water-Resistant, Packable, Navy Blue, XS-3X" matches specific queries much more precisely. The Google Merchant Center title guidelines are a useful baseline, the same specificity principles apply to ChatGPT's ACP feed.

How Do You Fix These Issues at Scale?

For stores with large catalogs, start with your top sellers by revenue. Fix the five problems above on your 50 highest-revenue products and measure the impact before rolling out to the full catalog.

For GTINs: if you manufacture your own products, you'll need to purchase a GS1 prefix. If you're reselling branded products, the GTIN should already exist, you just need to add it to each variant in Shopify.

For descriptions: the fastest approach is to build a template that forces you to answer four questions for every product: Who is this for? What problem does it solve? What are the key features/specs? What's the use case? Descriptions built on those answers will match AI queries because they're written the way buyers think, not the way catalogs are organized.

The feed is running. The question is whether it's working for you or just running.

Want to know which of your products are optimized for the ACP feed and which ones aren't? Check Your Store's AI Readiness →