Ad feed health
Google Shopping & Meta catalogs3 of 25 sampled products have an issue that gets items disapproved or throttled in Google Shopping — roughly 83 products across the 691.
Published audit · 16 October 2026
How Google Shopping and AI shopping agents read www.allbirds.com
We read Allbirds' storefront the way Google's crawler and agents like ChatGPT, Perplexity and Google AI Mode do — no login, no app install — and checked 25 flagship products of 691 against Merchant Center's rules and agent readability.
3 of 25 sampled products have an issue that gets items disapproved or throttled in Google Shopping — roughly 83 products across the 691.
10 of 25 are hard for AI shopping agents to read: key buying facts are missing or locked in prose and images — roughly 276 products catalog-wide.
Ready to use
The lowest-scoring products from the sample, rebuilt into complete structured data, Shopify metafields and Google feed rows.
Enter your work email to open the rebuilt products and download their JSON-LD, Shopify metafields and feed rows.
We'll email you this report's link. No newsletter; one follow-up at most. Privacy
What we found
Most severe first, with how many of the 25 sampled products each one affects.
Agents filter on facts ("merino, under 300g, machine-washable"). With none, the product can't match those queries.
How to fix Add the facts shoppers filter on — material, dimensions, weight, care — to the description and, better, to metafields published in the JSON-LD as additionalProperty.
Description is a required feed attribute; agents have nothing to reason over.
How to fix Write a description that covers what the product is, what it's made of, its fit or dimensions, and who it's for.
The identifiers are already there — the theme just doesn't publish them, so crawlers and agents never see them.
How to fix Add gtin to each variant's Offer in the theme's JSON-LD from variant.barcode — the data is already in Shopify.
Facts exist in paragraphs but not as attributes; agents and feeds have to guess, and usually drop them.
How to fix Move specs out of paragraphs into product metafields and publish them as additionalProperty name/value pairs in the JSON-LD.
Agents can't tell which sizes/colors exist or are in stock, so they can't answer "do you have it in M, in blue?"
How to fix Describe the product as a ProductGroup with hasVariant, or one Offer per variant, so each size and color carries its own price, availability, SKU and GTIN.
These are the attributes Google and AI agents filter on; they're absent from the machine-readable layer.
How to fix Publish color, material and pattern in the JSON-LD (the Product properties of the same names), filled from variant options or metafields.
Without a product type, mapping to Google's taxonomy (and bidding by category) falls back to guesswork.
How to fix Fill in the product type (or Shopify product category) for every product so it maps cleanly to Google's product taxonomy.
Short copy limits which searches the item can match in Shopping and gives agents nothing to quote.
How to fix Expand the description past 150 characters with material, dimensions, use and care — the facts agents quote and feeds match on.
Google requires these for apparel in key markets (US, UK, DE, FR, JP, BR), and agents filter on them. If they only live in app metafields, no crawler or agent can see them.
How to fix Publish color and size as variant options, and gender and age group as product metafields that both your JSON-LD and your feed read (Shopify's Google & YouTube app uses the mm-google-shopping namespace).
Google shows shipping and returns annotations only when they're marked up; agents compare stores on exactly this.
How to fix Add hasMerchantReturnPolicy and shippingDetails to the Offers in your JSON-LD, or configure return and shipping policies in Merchant Center.
Crawlers and agents that don't run vision read nothing from the images.
How to fix Add alt text to product images in Shopify admin, describing the product, its color and the view.
Product by product
Scores out of 100, worst first. Open a product to see what's wrong with it.
Store-wide
Enter your store and get both scores, every issue, and your five worst products rebuilt from your own photos and copy — free, in about a minute.
Audit your store free →Everything here comes from public storefront data: /products.json, the storefront Ajax API, each product page's JSON-LD, robots.txt and policy pages. Checks follow Google Merchant Center's product data specification and schema.org's Product vocabulary — see every check. Attributes kept only in app metafields or a separate feed aren't visible publicly — for AI agents, that's the same as missing.
Product fixes were extracted by anthropic/claude-haiku-5.5 from the store's own product photos and copy, and limited to facts those establish; anything it couldn't verify is listed as not determinable rather than guessed. Sample: 25 products — homepage picks first, then spread across product types. Generated 2026-10-16T09:45:44.110Z.