By Steve Merrill, Founder of WRKNG Digital — June 8, 2026
The 7 product feed fields that most affect ChatGPT's image-to-product matching are: primary image URL. Image alt text, color attribute, material attribute, style/type, additional image link, and Google Product Category (GPC). ChatGPT added image-upload shopping in May 2026 and now processes50 million shopping queries daily. These fields determine whether your product gets found — or missed entirely.
1. Primary Image URL (High-Resolution, Multiple Angles)
ChatGPT's visual search compares a user's uploaded photo against product images in your feed. The minimum threshold is 800x800 pixels — anything smaller fails the quality check for image-to-product matching. According toGoogle's product data specification, low-resolution images are one of the top reasons feeds get disapproved, and that same standard now applies to AI visual shopping.
2. Image Alt Text
Alt text is the text bridge between what the AI sees in an image and what your product actually is. Most Shopify stores default to just the product title — "Blue Shirt. " That's not enough.Descriptive alt text like "men's navy linen button-down shirt with chest pocket" gives the classifier multiple signals to confirm a visual match.Missing alt text means the AI is guessing.
3. Color Attribute
"Nice blue" won't match. "Navy blue" will. AI image classifiers map color terms against standardized color taxonomies, and vague names create mismatches between what a user sees in a photo and what your feed says the product is.Google's color attribute specificationallows up to three specific color values per product — use all three when they apply.
4. Material / Fabric Attribute
Texture is one of the strongest signals in image recognition. A linen blazer, a velvet cushion, a leather tote — each has a distinct visual texture the AI uses to narrow a match. If your feed says "fabric" instead of "100% cotton" or "genuine leather," you give the classifier nothing to confirm the match with. Specific wins every time.
5. Style / Type Attribute (Product Category Precision)
"Shoes" is not a product type. "Slip-on canvas sneaker" is. OpenAI'sAgentic Commerce Protocol (ACP)transmits product data — including type attributes — from your feed directly to ChatGPT's shopping layer. The tighter your style/type field, the more precisely the image match resolves to the correct product. Broad product types create broad, unreliable results.
6. Additional Image Link
Most Shopify stores submit one image per product. That's one angle. One shot at matching a user's photo.Google Merchant Center's additional_image_link fieldaccepts up to 10 supplementary images per product — different angles, detail shots, lifestyle images. A user uploads a photo of a product from the side, and your feed only has a front-facing shot. That's a miss you didn't have to take.
7. Google Product Category (GPC)
Google's product taxonomy has over 6,500 categories. Visual classifiers use GPC to organize product types before they match images. If you leave GPC blank or pick a top-level bucket like "Apparel," you're asking the classifier to do extra work — and it often gets the placement wrong. A specific GPC like "Apparel & Accessories > Clothing > Shirts & Tops > Button-Down Shirts" puts your product exactly where the image search expects it to be.
How We Chose This List
These seven fields come from cross-referencing Google Merchant Center's feed specification. OpenAI's ACP documentation, and feed audits run through WRKNG Digital's AI readiness assessment across live Shopify stores. Google also expanded visual shopping in AI Mode in early 2026, which means these same fields now affect two major AI platforms simultaneously. This isn't theoretical. These are the gaps showing up in real feeds right now.
FAQ
Does ChatGPT use my Google Shopping feed for image search?
Not directly — but the data flows through ACP from your product feed to OpenAI's shopping layer. If your feed is clean for Google's visual shopping, you're most of the way there for ChatGPT too.
How many images should I submit per product?
As many as you have, up to the platform limit. Google Merchant Center allows up to 10 additional images per product. More angles means more opportunities for a visual match to succeed.
What happens if my image alt text is missing?
The AI has no text context to confirm what the image shows. That makes image-to-product matching less accurate. Every product image should have descriptive alt text that includes color, material, and style — the product name.
Is Google Product Category mandatory?
Google marks GPC as optional in their spec. In practice, feeds without GPC data score lower on AI readiness audits and surface less often in AI-driven shopping results. Optional on paper, costly in practice.
Does this apply to all Shopify stores or just those running Google Shopping ads?
Every Shopify store that wants AI visibility — ChatGPT shopping, Google AI Mode, Perplexity product results — needs these fields clean. Visual shopping is live across all major AI platforms now, Google.
If you want to see exactly where your feed is falling short — image quality, missing attributes. GPC gaps — the AI readiness audit at WRKNG Digital shows you the full picture with specific fixes, not a generic checklist.
