[how-to]
AI getting your product wrong? A fix guide for Shopify
When ChatGPT or another AI agent states something false about your product, here’s how to trace where it came from and what you can change.
The Hearsay team Sep 30, 2026 6 min read
A customer emails to say ChatGPT told them your sunscreen is safe from birth, and your label says to ask a pediatrician under six months. Or an AI answer lists your blanket as machine washable when it’s dry clean only. The instinct is to rewrite your product page. Sometimes that helps. Often the claim didn’t come from your product page at all.
This guide is about finding the source first, because each source has a different fix.
Step 1: write down exactly what was said
Before you change anything, capture the claim:
- The exact wording, in quotes.
- The question that produced it, if you know it.
- Which AI channel, and the date.
AI answers vary from one question to the next, and product data changes. If you fix something and the answer changes, you’ll want to know which change did it. If you end up contacting Shopify or the channel, you’ll need the exact text.
Step 2: find where the claim came from
For a Shopify store, a wrong claim in an AI shopping answer usually traces back to one of four places.
Source A: your own copy
Check this first, because it’s the most fixable. Search your description, product details fields, metafields, image alt text and product page for the claim or anything close to it. Look for:
- Old copy. A claim from a previous formula or a previous supplier that never got removed.
- Copy that disagrees with itself. A description that says “gentle for all ages” with an ingredients tab that says “not for use under 6 months.”
- Claims in images only. A “fragrance-free” badge in a product photo with nothing in the text.
Source B: Shopify Catalog’s inferred fields
For most AI channels, Shopify stores’ products arrive through Shopify Catalog, and the catalog doesn’t just pass your copy along. Shopify’s developer docs mark the catalog’s description, top features, tech specs and selling points as Inferred: “generated or enriched by Shopify,” and they “might vary in accuracy depending on available product data” (Global Catalog docs).
That means an agent can be told something about your product that appears nowhere in your copy. We explain each inferred field in What Shopify’s AI adds to your listings.
This happens more than you’d expect. When we read the catalog’s version of 192 top-ranked products and checked 3,546 claims against each brand’s own copy, 89% of the products had at least one claim the brand never made. 70% had one about health, safety, certification or what the product does (how we checked). The examples are in the study.
Source C: the AI channel’s own rewrite
Channels can reword what they receive. OpenAI says “ChatGPT may generate simplified product titles and descriptions based on information it receives from third-party providers” (Shopping with ChatGPT search). A claim can be right in the catalog and still come out wrong in the answer.
Source D: the open web
AI channels also read reviews, articles and other pages. OpenAI says its review summaries “are based on reviews from public websites” and that “reviews and ratings are not verified by OpenAI.” An old retailer listing or a confused review can end up in an answer.
For Google’s AI Mode and Gemini, add one more place to check: Shopify says those channels get products through the Google & YouTube channel or Google Merchant Center (Selling on Google AI Mode and Gemini). If your Merchant Center data is stale, that’s your source.
Step 3: fix what you control
If the source is your copy
Fix it at the source and make it consistent everywhere: description, metafields, product details, tabs and images. Then make the true version hard to miss:
- State important facts in plain words, early. If the product contains fragrance, say so in the description, not only in an ingredient list three tabs down.
- Keep required disclosures near the top. Shopify tells merchants selling through ChatGPT, Copilot, Google and Meta to “ensure that your product descriptions include any relevant legal disclosures in the first 6,000 characters” (Selling on ChatGPT).
- Be exact with numbers and ages. “Ages 4 and up” is safer than “for kids,” and an SPF, size or count should be the same everywhere it appears.
If the source is the catalog
You can’t edit the catalog’s inferred text. Nobody outside Shopify can. You can change what it’s built from:
- Make the fact explicit in your own copy, as above. A summary has less room to go wrong when the source is clear.
- Check Catalog Mapping under Sales channels > Agentic. If your real product data lives in metafields, you can point the catalog’s title, description and category at them. Shopify says the mapping settings “are inputs that can influence Shopify Catalog” and “changes are processed with a delay” (Mapping your product data sources).
- If your copy is already clear and the catalog still says otherwise, contact Shopify Support with both versions side by side. There’s no published process for correcting an inferred claim, so be specific about which product, which field and which words.
If the source is the channel or the web
- Store-level questions like returns, shipping and policies can be addressed in the free Shopify Knowledge Base app, which lets you customize the FAQs AI platforms use to answer questions about your store (Shopify Knowledge Base). It doesn’t edit individual product claims.
- Google: bring your Merchant Center data in line with your store.
- Stale third-party pages: ask the site to update, like you would for any other outdated listing.
Step 4: check again, and give it time
Changes don’t show up instantly. Shopify says Catalog Mapping changes are processed with a delay and doesn’t say how long. When you edit a product, wait a few days before judging whether the catalog caught up, and then check the same claim again.
If you only remember one step: write down the claim with the date, so you can tell whether a later change helped.
Common questions
Is this worth the effort?
It depends on the claim. A wrong color name is an annoyance. A wrong age range, allergen or certification is a customer-safety problem and a returns problem. In our study, 14% of 192 top-ranked products were described to AI agents with a claim that directly contradicts the brand on health, safety or what’s in it. For baby products it was 1 in 5 (how we checked).
Can I stop AI channels from describing my products at all?
You can turn off Shopify Catalog access for ChatGPT, Copilot and other third-party channels, or hide a product with the Unlisted status. Shopify warns that products can still be found “through web crawling and indexing,” and that Unlisted also hides the product from search engines and your own store search (Shopify Catalog and product discovery). For most stores that costs more than it saves.
How do I catch this before a customer does?
Check the catalog’s version of your products on a schedule, not after a complaint. That’s what Hearsay does. It reads what Shopify Catalog tells AI agents about each product, compares every claim with your own copy, and emails you when a claim changes. Each finding comes with a note on what you could change in your own listing (Fixing a finding). The free scan checks 5 products, and the app’s Free plan rechecks your 5 most recently updated products every day.