By Steve Merrill, Founder of WRKNG Digital — June 18, 2026
How Does ChatGPT Actually Decide Which Products to Recommend?
Nobody knows the full algorithm. But after auditing 40+ Shopify stores, I can tell you what the data shows.
The stores showing up in ChatGPT Shopping, Perplexity's product results, and Google AI Overviews share a common trait. It's not ad spend. It's not domain authority. It's signal quality , the completeness and accuracy of the data they're sending to AI systems.
Most Shopify stores are sending bad signals. Not because store owners are careless. Because nobody told them that the rules changed.
Structured Data Is the Price of Entry
This is non-negotiable. If your product pages don't have complete schema.org Product markup, you're effectively invisible to most AI recommendation engines.
"Complete" doesn't mean a title and a price. It means:
- Product name, description, and images
- Price and currency , current, not cached from six months ago
- Availability status (in stock, out of stock, pre-order)
- Brand and SKU
- Aggregate rating and review count
- Product category using standardized Google Product Taxonomy values
I've audited stores with 300+ products where fewer than 15% had complete structured data. The rest were either missing required fields or feeding outdated information. Those stores weren't showing up in AI results. Not because AI doesn't know they exist , because AI can't confidently recommend a product when the data is incomplete or contradictory.
AI systems don't guess. They recommend what they can verify.
Google's Product structured data documentation outlines exactly which fields are required versus recommended. Most Shopify themes handle the required fields automatically. The recommended fields are where most stores fall apart , and those recommended fields are what separates your product from three competitors when an AI has to choose one to surface.
Reviews Are a Ranking Signal Now
Every AI shopping assistant I've tested weights review signals heavily.
Not just the star rating. The full picture: total review count, rating distribution, recency of recent reviews, and whether the reviews contain specific language about the product's use case and quality.
Here's what I've observed in practice. Products with fewer than 20 reviews almost never show up in AI recommendations, even with perfect structured data and competitive pricing. Products with 50+ reviews and a rating above 4.2 show up consistently. Products with 100+ reviews and strong recent velocity , those are the ones getting recommended first.
Read that again. Review count matters as much as the rating itself.
There's a reason for this. AI systems are trying to minimize bad recommendations. A product with 500 reviews at 4.1 stars is lower risk than a product with 8 reviews at 4.9 stars. The signal is more reliable. The confidence is higher.
One thing most stores miss: your reviews need to be marked up in your structured data. If you have 200 reviews but your schema reports only 12 because it was last updated six months ago, AI is reading the 12. Not the 200. That's a fixable problem that directly affects whether you get recommended.
Price Positioning Matters More Than Price Level
AI shopping assistants compare. Full stop.
When someone asks ChatGPT "what's a good [product category] under $100," it's pulling from multiple sources and evaluating your price against product quality signals simultaneously. You don't have to be the cheapest option. But your price needs to make sense relative to what you're offering.
What kills stores in this analysis isn't being expensive. It's being expensive with no supporting signals. High price plus weak reviews plus thin product description equals no recommendation. High price plus strong reviews plus a detailed product description plus real brand presence? That can get recommended , often as the premium pick.
The other issue is price accuracy. If your product feed says $79 but your product page says $89 because you ran a sale six months ago and forgot to update one of them, AI systems catch that discrepancy. It creates a trust problem. The product gets deprioritized.
Inconsistent data is the fastest way to get ignored by AI.
Brand Authority Across the Web
AI doesn't only look at your product pages. It looks at everything about your brand that exists on the web.
Third-party mentions. Product reviews on sites outside your store. Press coverage. Social proof. The web footprint your brand has built over time. This is ambient authority , the signal that comes from existing in multiple places, not just your own domain.
I've run audits on stores that had perfect structured data, strong reviews, and competitive pricing , and still weren't showing up in AI recommendations. When we dug into the brand's web presence, the answer was clear every time: almost no external mentions. No editorial coverage. No product reviews outside their own store. To AI, they looked like a brand that only existed on its own website. That's low-confidence territory.
The fix isn't complicated. Get your products in front of creators who write actual reviews. Pitch coverage to niche publications in your category. Build a brand presence that exists outside your own domain. Not for backlinks , for AI trust signals. There's a difference, and it matters.
Your Product Descriptions Are Doing More Work Than You Think
Here's where most Shopify operators are leaving the most visibility on the table.
When someone asks an AI assistant "what's the best yoga mat for bad knees," the AI is scanning product descriptions, not just titles and prices. It's looking for content that directly answers that specific question.
Most product descriptions don't answer anything. They describe. "This yoga mat is made from premium TPE foam with a non-slip surface and measures 72 x 24 inches." Fine for a product listing. Useless for an AI recommendation query.
The stores getting recommended have descriptions that read more like: "Designed for joint support , the 6mm thick TPE base distributes pressure evenly, making it a strong choice for practitioners with knee or hip concerns. Tested by physical therapists. Works on hardwood, carpet, and gym floors."
Same product. Completely different signal to AI.
Shopify covers the basics of product content in their SEO documentation. The AI layer takes it further. You're writing for a system that's trying to match your product to a specific human need. If your description doesn't answer the question someone is actually asking, you won't get the recommendation , even if your product is the right answer.
What This All Adds Up To
Most Shopify stores aren't losing AI visibility because AI doesn't like them. They're losing it because they haven't given AI what it needs to confidently recommend them.
Complete structured data. Real review volume. Accurate pricing. Third-party web presence. Product descriptions written for humans asking specific questions. These aren't advanced tactics. They're fundamentals most stores skip because nobody told them AI was watching.
I've audited enough stores now to see the same pattern repeat. The ones getting recommended aren't always the biggest brands or the best products. They're the ones whose data is clean, complete, and consistent across every surface AI checks.
That's a solvable problem. For most stores, it's a 60-90 day fix.
The window to do this before your competitors figure it out is still open. Not forever.
Frequently Asked Questions
How does ChatGPT decide which products to recommend?
ChatGPT uses a combination of structured data signals (schema.org Product markup), review quality and quantity, price competitiveness, brand authority across the web, and the completeness of your product information. Stores with complete, accurate structured data and strong review bases get recommended far more often than stores without them.
Does my Shopify store need special technical setup to show up in AI recommendations?
Yes. You need proper schema.org Product markup on every product page, an accurate product feed, and strong review signals. Most default Shopify themes include basic structured data, but "basic" is not enough for AI recommendations. You need complete markup including availability, pricing, brand, and review aggregates.
What is the minimum review count needed for AI product recommendations?
There is no published minimum, but from auditing Shopify stores, products with fewer than 20 reviews rarely appear in AI shopping recommendations. Products with 50+ reviews and a rating above 4.2 stars show up consistently across ChatGPT, Perplexity, and Google AI Overviews.
Does price affect whether AI recommends my product?
Yes. AI shopping assistants actively compare prices across sources before recommending. This doesn't mean you need to be cheapest , it means your price needs to be clearly communicated, accurate in your product feed, and reasonable relative to what competitors charge for similar items.
How long does it take to see results after fixing structured data for AI shopping?
Most stores see changes in AI visibility within 4-8 weeks of fixing structured data issues, assuming AI models recrawl their products during that window. Review signals take longer. Building a strong review base is a 3-6 month process for most stores.
Find Out Where Your Store Stands
If you're not sure whether your Shopify store is sending the right signals to AI shopping assistants, that's the first thing to fix. We built an audit process specifically for this , checking structured data completeness, review signal health, product feed accuracy, and brand authority across the sources AI actually uses.
