By Steve Merrill — September 21, 2026
What happens when you ask 5 AI assistants to recommend a product?
You get five overlapping answers, not one. I asked ChatGPT, Perplexity, Google AI Overviews, Gemini, and Microsoft Copilot to recommend a product in one category. The same handful of brands kept showing up. The reason was rarely product quality. It was who had clean data, real reviews, and coverage the models could cite.
I run WRKNG Digital. I spend my days looking at why AI assistants name some stores and ignore others. So I stopped guessing and ran the test myself.
How did I run the test?
I picked one product category and wrote one plain question. Something a normal shopper would type: "What's the best [product] for someone on a budget?" Then I asked all five assistants the exact same thing. Same wording. Same day.
I wasn't chasing precise numbers here. This was a representative test, not a lab study. I wanted to see the pattern, not publish a spreadsheet. And the pattern showed up fast.
Across the five tools, I saw maybe eight or nine brands named in total. But only three or four showed up in more than one answer. A small group did the heavy lifting. Everyone else got a mention once, or never.
Which brands got recommended, and why?
The brands that showed up everywhere had one thing in common. They were easy to read and easy to verify. Each one had structured product data on its pages, a stack of reviews across multiple sites, and at least one write-up on a source the models already trust. The product being good was table stakes. It wasn't the deciding factor.
Here's what I noticed across the winners.
Structured data was the price of entry. The brands that got named had clean product markup. Price, availability, specs, and review counts were all readable by a machine, not buried in an image or a PDF. Google's own guidance on product structured data spells out why this matters. If the model can't parse your page, it can't quote your page.
Reviews came up again and again. The recommended brands had review volume that stretched across their own site plus outside platforms. That gives an assistant something to cite. One brand I love in this category makes a genuinely better product. It got skipped in four out of five answers. Why? Barely any reviews. The model had nothing to hang its recommendation on.
Third-party citations sealed it. Perplexity in particular leaned hard on review roundups and "best of" lists. Perplexity shows its sources right in the answer, so I could see exactly where each pick came from. Brands with press coverage and independent reviews got pulled in. Brands with only a homepage got left out.
Why did the same brands not win everywhere?
Because each assistant reads from a different pile of sources. That surprised me at first. Then it made sense.
Google AI Overviews leaned on Google's own index and shopping data. Gemini pulled from a similar place but phrased things differently. ChatGPT with search pulled a wider mix, including some smaller sites. Copilot sat somewhere in the middle. Perplexity was the most transparent about where it looked.
So the answers overlapped in the middle and split at the edges. The brands with the widest footprint were the only ones that showed up across all five. If your store only lives in one place, you only show up in one place.
What about feed freshness?
Freshness mattered more than I expected. One brand got recommended with an old price that was flat wrong. The assistant pulled stale data because the feed hadn't updated. That's a trust problem waiting to happen. A shopper clicks through, sees a different price, and bounces.
Stores that keep their product feed current give the models accurate, confident answers. Stale feeds create doubt, and doubt gets you dropped. Shopify stores that sync their catalog often have a real edge here, because the data the assistant reads actually matches the store.
What's the lesson for Shopify stores?
The lesson is simple, and a little uncomfortable. Having the best product doesn't get you recommended. Being the most readable and most verified brand does. AI assistants recommend what they can confirm.
Here's the bottom line. If you want to show up when a shopper asks ChatGPT or Perplexity for a recommendation, you need three things working together. Clean structured data on every product page. Real reviews across more than one site. And coverage on sources the models already cite.
I've seen this pattern in audit after audit. The store owner swears their product is best in class. And it might be. But the AI can't taste the product. It can only read the data. Fix the data, and you give yourself a shot at the answer.
Start with your product pages. Make sure price, availability, specs, and reviews are all in structured format a machine can read. Then get your feed syncing on a real schedule. Then go earn a few honest reviews and one solid third-party mention. That's the order I'd run it in.
Most stores skip all three. That's your opening.
Frequently asked questions
Why do AI assistants recommend some brands and skip others?
AI assistants pull from what they can read and verify. Structured product data, third-party reviews, and citations from sources they trust. Brands with clean feeds and outside coverage get named. Brands with thin data get skipped, even when their product is better.
Does product review count affect AI recommendations?
Yes. In my test, the recommended brands had steady review volume across multiple sites. Reviews give the model social proof it can cite. A great product with 11 reviews rarely beats an average product with a few thousand.
How do I get my Shopify store recommended by ChatGPT and Perplexity?
Fix your structured data first, keep your product feed fresh, and earn coverage on sites AI models already cite. Add real reviews and clear specs to every product page. The goal is to be easy to read and easy to verify.
Do all AI assistants recommend the same brands?
No. Each assistant pulls from different sources, so the answers overlap but don't match. Perplexity leaned on review roundups. Google AI Overviews leaned on its own index. The brands that showed up everywhere had the widest footprint.
Is this test something I can run for my own store?
Yes, and you should. Ask all five assistants to recommend a product in your category. See who gets named. Then look at what those brands have that you don't. It's the fastest way to find your gaps.
Want help showing up in AI recommendations?
We help Shopify stores get read and recommended by AI assistants. If you want to see where your store stands and what's holding it back, take a look at how we run agentic commerce audits.
