← Back to all posts
Shopify AI search: what your product feed must contain

Shopify AI search: what your product feed must contain

LA
Jedi Solomon
··5 min read

Shopify AI search starts with your product catalog. Before ChatGPT, Gemini or Perplexity weighs anything on your pages, it needs a product record it can trust. Shopify's catalog needs a standard category, an image for every variant, a price above zero and complete store policies. A product that fails those checks is a weak candidate for any assistant. Locus found that 45% of US online shoppers use an AI tool somewhere in the purchase journey (Q2 2026), so assistants are already part of how many buyers shop.

Why the feed deserves your first week

The share is still small, and it is growing fast. First Page Sage reports that AI-referred visits were 4.73% of US retail site visits in Q3 2026, up from 0.43% in Q4 2024, and that they converted about 1.6x other traffic. Adobe measured AI traffic to US retailers up 393% year over year in Q1 2026, converting 42% better than non-AI traffic. Shopify said AI-driven traffic to its stores rose 8x in the same quarter, and AI-search orders rose about 13x. The three sources measure different things, so the figures do not line up exactly, and they point the same way.

What the catalog must contain

Shopify's catalog has four hard requirements:

  • A standard product category, chosen from Shopify's own taxonomy.
  • An image for every variant, so the black size 10 has its own picture.
  • A price above zero on every product and variant.
  • Full policies: shipping, returns and refunds, written out and findable.

If you want your products in ChatGPT shopping, OpenAI's product feed spec adds its own list. It has 9 required fields. Pull the spec, map each field to where it lives in Shopify, and look for products where that field is empty.

Fixes you can make this week

  1. Export all products from the Shopify admin as a CSV and sort by the category column. Fill every blank with a standard category. With thousands of SKUs, do bestsellers first and the long tail next.
  2. Filter for variants with no image. Assign one to each. A color variant that shows the wrong photo is worse than one with a plain photo of the right color.
  3. Search for price 0. Look for samples and old placeholders. Fix the price or unpublish the product.
  4. Read your policy pages as a stranger would. Look for shipping times, countries served, return window, who pays return shipping and how long refunds take. Fill every gap in plain sentences.
  5. Rewrite titles and descriptions the way a buyer asks. "Women's linen shirt, relaxed fit, sizes XS to 3X" gives an assistant something to match. "Summer Edit 04" gives it nothing.

Assistants answer questions like "a linen shirt that does not cling, under $80, ships in three days." Every attribute in that sentence has to exist somewhere in your record.

What to leave alone

Skip llms.txt. In May 2026, 97% of llms.txt files received zero AI crawler requests, and Google discourages the file.

Structured data correlates only weakly with AI citations. It still matters for rich results and for feed consistency, so keep price and availability in your markup identical to what the catalog says.

Shopify AI search also reads your pages

The feed covers the product record. Questions that come before the purchase get answered from pages: how to size a jacket or which fabric suits humid weather. If you have no page that answers them, an assistant uses someone else's.

Ahrefs found that 38% of AI Overview citations come from the top 10 results and 31% from positions 11 to 100 (March 2026). Pages that rank modestly can still be cited. The same Ahrefs study of 75,000 brands put the correlation of third-party brand mentions with AI Overview visibility at 0.66, against 0.22 for backlinks. Reviews and forum threads that name your brand are worth chasing before another batch of links.

JCS Fashions is a 4,878-product Shopify apparel store in Milpitas, California, and LynkAI's first customer. In 2026 we cleaned up its catalog data and published about 290 quality-gated articles between April and September. Guide-page clicks went from 0 in February to 231 in August, and 233 guide pages earned impressions on 24,164 queries. Sessions from AI assistants went from 20 in January to 569 in October. Both pieces of work ran in the same months, so we cannot split the credit between them. The full customer story has the monthly numbers, and the ecommerce page describes how we handle catalogs and feeds.

Check where you stand

Start with the CSV export. Count the products missing a category, the variants without images and the items priced at zero. That count tells you how much of your catalog AI shopping assistants can use today. When you want to see how your store shows up for real buyer questions in ChatGPT, Gemini and Perplexity, run the free AI visibility scan.