By Steve Merrill, Founder of WRKNG Digital — July 6, 2026
Three seconds. That's roughly how long it takes ChatGPT to go from reading a product question to naming the specific item it recommends, the price, and the reason why. Not much time. But a lot happens in it.
I've pulled apart probably 60 ChatGPT Shopping recommendations for client stores over the last few months, comparing what showed up against what didn't. Same category, similar products, wildly different outcomes. The pattern isn't mysterious once you see it enough times. It's a data problem, not a luck problem.
Here's the walkthrough of what actually happens in that window, and what it means if you run a Shopify store.
What Does ChatGPT Check First When Someone Asks for a Product?
The first pass is structured data, not your homepage copy. ChatGPT Shopping pulls from product feeds (the same kind of feed structure Google uses for Google Merchant Center) to build a fast shortlist: title, category, price, brand, GTIN or MPN, and availability status.
This is the filtering step. If your feed is missing a GTIN, has a vague title like "Blue Shirt" instead of "Men's Organic Cotton Crew Neck T-Shirt, Navy," or lists a category that doesn't match how people actually search, you get filtered out before the model ever looks at your page.
Most stores lose here. Not because their products are bad. Because their feed data is thin.
How Does Schema Markup Change Whether a Product Gets Recommended?
Once a product survives the feed filter, ChatGPT cross-checks it against the actual page using structured data, specifically Product schema. This is where a lot of otherwise good products fall out of contention.
The model is looking for agreement between what the feed claims and what the page's schema confirms: price, currency, availability, condition, and review aggregate. When those match, the product gets treated as verified. When they don't, the AI has a choice: trust the feed, trust the page, or skip the product to avoid recommending something wrong. In my experience, it skips it.
Shopify's default themes don't always render complete Product schema out of the box. You can check yours in a couple minutes using Google's Rich Results Test. If it comes back with warnings or missing fields, that's a direct line to lost recommendations.
The Fields That Matter Most
Not every schema field carries equal weight. From what I've seen across audits, these five get checked hardest:
- Price and priceCurrency (must match the feed exactly)
- availability (InStock, OutOfStock, PreOrder — and it has to be current)
- aggregateRating and reviewCount
- brand
- gtin13 or mpn
Miss two of these and your odds of showing up drop hard.
Why Do Reviews Matter So Much in the Final Seconds?
Here's the thing. When ChatGPT has three similar products that all pass the feed and schema checks, it needs a tiebreaker. Reviews are usually it.
A product with 340 reviews at 4.6 stars reads as a safer recommendation than one with 6 reviews at 4.9 stars, even though the second one has a "better" rating on paper. Volume signals real purchase history. The model is optimizing for a recommendation the user won't regret, and review count is the cheapest proxy for that.
We ran this exact comparison for a client's supplement store last month. Two products, same ingredients, same price point. The one with 210 reviews got recommended in 8 out of 10 test prompts. The one with 12 reviews got recommended once. Same everything else.
If your review count is thin, that's not a copywriting problem. It's a math problem, and it takes time to fix. Start collecting them now.
What Role Does Page Content Play Beyond the Schema?
Schema gets you into consideration. Page content is what convinces the model your product actually solves the person's problem.
ChatGPT reads the visible product description, the specs table, and often the FAQ block on your page to match language against the user's original question. If someone asks for "a waterproof phone case for kayaking" and your description only says "durable case, great protection," you're giving the model nothing specific to latch onto. Specificity wins here. Waterproof rating, drop-test height, material, actual use cases. This is also where a lot of Shopify product pages default to whatever the supplier's boilerplate description was. Copy-pasted, generic, no mention of the exact use case a real buyer would type into ChatGPT. Fixing this is slow work but it's the biggest thing most stores haven't touched.
What Happens If Your Data Doesn't Line Up?
Not great. When the feed, the schema, and the page content disagree, the product usually just gets left out. ChatGPT doesn't flag it for a human to fix. It moves on to the next candidate. This is the part that surprises store owners most. They assume a bad recommendation means their product ranked low. Usually it means their product never made the shortlist at all, because one data source contradicted another.
What Should a Shopify Store Owner Actually Do About This?
Four things, in order of impact:
- Audit your Shopify product feed for missing GTINs, vague titles, and stale availability status.
- Check Product schema rendering with Google's Rich Results Test on your top 20 SKUs.
- Build a simple review collection flow if you're under 20 reviews per product (Shopify's own product merchandising documentation covers the technical setup for review integrations).
- Rewrite product descriptions with specific use cases instead of generic adjectives.
I've seen this exact pattern in 40+ audits now. Stores that fix the feed and schema first, before touching ad spend or content volume, see AI recommendation visibility move within weeks. Stores that skip straight to content without fixing the data underneath stay invisible no matter how much they write.
Frequently Asked Questions
Does ChatGPT Shopping use my Shopify product feed or crawl my site directly?
Both. ChatGPT Shopping leans on structured product feed data (title, price, availability, GTIN) for the fast match, then checks your live page for schema markup and content to confirm the details before recommending it.
Do I need a paid app to show up in ChatGPT Shopping?
No. Clean Product schema, an accurate feed, and honest availability data matter more than any app. Some feed management apps help, but they don't replace fixing the underlying data.
How much do reviews actually affect whether ChatGPT recommends a product?
A lot. Review count and rating are two of the strongest signals ChatGPT uses to break ties between similar products. Products with zero reviews or fake-looking review clusters get skipped more often.
What breaks a product's chance of being recommended fastest?
Availability mismatches. If your feed says in stock but the page shows sold out (or the reverse), that product usually gets dropped from consideration entirely rather than flagged for review.
Can I test what ChatGPT sees before it recommends my product?
Yes. Pull your product page through Google's Rich Results Test to check your schema, then ask ChatGPT directly about your product category and see whether you show up and what details it cites.
Want This Fixed for You?
Auditing your own feed and schema takes a few hours if you know exactly what to look for. Most stores don't, and they burn a weekend chasing the wrong fix. We do this for a living. If you want your Shopify store checked against everything ChatGPT actually looks at before it recommends a product, start here: wrkngdigital.com/agentic-commerce-landing-page.
