We checked 11 smart pet brands' product pages. Here's why AI shopping assistants can't see them clearly
When AI recommends products, it compares fields that machines can read. We checked 11 smart pet brands, mostly from China, against OpenAI's published product feed spec.
More and more US shoppers ask ChatGPT or Google's AI before they buy. Adobe data show that generative AI traffic to US retail sites grew 693% year over year in the 2025 holiday season, and those visitors were more likely to buy than other traffic (source).
So what does AI actually look at when it recommends a product?
AI doesn't read ad copy. It reads fields
OpenAI has published the product feed spec used for shopping in ChatGPT. Required fields are title, description, link, brand, seller, image, availability and price. Optional fields include GTIN, materials and dimensions, shipping, return policy, product reviews and store reviews (OpenAI spec).
In other words, these fields are what AI compares. When a field is missing, the AI skips it, looks elsewhere, or guesses.
What we found on 11 product pages
We picked smart pet products: automatic feeders, water fountains and smart litter boxes. It is a spec-heavy category that buyers love to ask AI about, and many of the brands are from China. We took one product page from each brand's own site and checked whether its structured data (the part of a page written for machines) included these items:
- Product barcode (GTIN)3 / 11
- Product-level shipping info1 / 11
- Return policy markup3 / 11
- Rating markup5 / 11
The one brand with product-level shipping info was not a Chinese brand. We also found more basic issues: variant weights set to 0, empty SKUs, descriptions of a few dozen words with all the specs inside images, Chinese punctuation in English titles, and an About page that gave two different founding years.
Limits: one page per brand, page markup only, no access to what brands submit to Google Merchant Center. This doesn't mean these brands do badly overall, but it shows how common these gaps are.
Why small gaps matter
US buyers don't ask AI the way they type into a search box. They ask: “I have two cats and one is on a prescription diet. Which feeder can feed them separately?” “Will the feeder still work if the power goes out while I'm away for three days?” “Who pays for return shipping?”
Adjectives like high quality, smart or best can't answer those. Only specific specs and policies can. If the product page doesn't say it, AI can't match your product to the question.
Worse, AI sometimes fills the gap itself. No delivery time? It may say two days. No return policy? It may say no returns. Wrong descriptions turn into bad reviews and refunds.
Five things to do first
- Check robots.txt and your CDN settings so you are not blocking crawlers from ChatGPT, Perplexity and other AI services.
- Write shipping, delivery time, return window and warranty as clear, separate policy pages.
- For each product, fill in the 6 to 10 specs buyers compare most, with real numbers and local units, not only inside images.
- Get your own GTINs if you can, and fill in weight and SKU for every variant.
- Say who the product is for, when to use it and when not to, so the description answers what buyers will ask.
Finally: don't trust a recommendation guarantee
AI recommendations change every time. One study found that two runs of the same prompt return the same list of brands less than 1% of the time (source). Any promise that ChatGPT will recommend you is not credible. What you can do is make product information complete and accurate, which raises the odds of being picked.
Our test method is public: see Method.