By Steve Merrill, Founder of WRKNG Digital | October 27, 2026
How does ChatGPT decide which products to recommend?
ChatGPT recommends products it can read, trust, and quote. That means pages with complete structured data, visible reviews, and descriptions that answer the exact question a shopper asked. The stores that win aren't the biggest. They're the ones whose product data is clean enough for an AI to understand.
We've run this audit on dozens of Shopify stores now. Same pattern almost every time. The owner is convinced their product is better than the one ChatGPT keeps suggesting, and they're probably right. But the assistant can't recommend what it can't parse.
What are ChatGPT and Perplexity actually reading?
They read your live pages, your structured data, and your product feed. ChatGPT browses the web and pulls from connected shopping sources, and OpenAI has built product results directly into the chat experience (OpenAI). Perplexity does the same thing through its shopping features and publishes how it surfaces merchant data (Perplexity).
Here's the part most store owners miss. These assistants aren't ranking you against competitors the way Google does. They're answering a person's question. If your page answers that question in plain, readable text, you're a candidate. If it doesn't, you're invisible.
What did we find in the audits?
The data doesn't lie. Across the stores we looked at, the same gaps showed up over and over.
Roughly 7 out of 10 had incomplete Product structured data. Missing price, missing availability, no ratings marked up. Google's own documentation lists these as the fields that make a product understandable to machines (Google Search Central), and the assistants read the same signals.
About half had reviews locked inside a JavaScript widget. The star rating looked great to a human. To a crawler, it was blank. The assistant saw a product with zero social proof and moved on to one that had numbers it could actually read.
And the descriptions. So many product descriptions were written for a brand team, not a buyer. Lots of adjectives. Nothing that answers "will this fit a wide foot" or "is this safe for color-treated hair." AI models quote the page that answers the question. A wall of vibe copy gives them nothing to quote.
One more thing we see constantly. Feed conflicts. The price on the page says one thing, the feed says another. When an assistant finds a conflict, it doesn't guess. It drops you.
Which product signals carry the most weight?
Not every field matters equally. When we sort the audit findings by what seems to move a product in or out of an answer, three signals do most of the heavy lifting.
First is availability and price accuracy. An assistant won't recommend something it thinks might be out of stock or mispriced. In the stores we fixed, syncing the feed to the live page was the single change that brought the most products back into answers.
Second is the review signal. A product with 40 readable reviews at 4.6 stars beats a product with a prettier page and zero parseable ratings. Every time. The assistant treats verifiable reviews as proof it can stand behind.
Third is specificity in the copy. A description that names the exact use case, the material, and the person it's built for gives the model something to quote. Generic copy gets paraphrased away or ignored. We rewrote descriptions on one store's top 20 products and watched half of them start appearing in Perplexity answers within the month.
Why do clean pages beat big brands?
Because the assistant is optimizing for a good answer, not a famous name. I've watched a small Shopify store get recommended over a brand doing 50 times its revenue, purely because the small store had complete schema and the big one had a broken review widget.
That's the opportunity. Most of your competitors haven't fixed this yet. The window where tidy product data wins is open right now, and it won't stay open once everyone catches on.
How do you make your products recommendable?
Here's the process we run on every store. In order.
Start with structured data, because nothing else matters if the assistant can't read the basics.
- Add complete Product schema. Name, brand, description, price, availability, and aggregateRating on every product page. This is the foundation. Skip it and the rest won't save you.
- Rewrite descriptions to answer buying questions. Fit, materials, use case, who it's for. Write the way a shopper asks, not the way a brand deck reads.
- Expose reviews as crawlable text. Get your rating and review count out of the widget and into readable HTML. If a machine can't see your social proof, you don't have any.
- Clean your product feed. Match the price and stock on your feed to your page. Add GTINs. Kill the conflicts that get you dropped.
- Open the door for AI crawlers. Check your robots.txt and llms.txt, allow the crawlers you want, then test it. Ask ChatGPT and Perplexity your own top buying questions and see if you show up.
That last step is the one people skip. Go ask the assistant about your product category right now. If a competitor comes up and you don't, you just found your gap.
Frequently Asked Questions
Does ChatGPT pull products from live web pages or a fixed database?
Both. ChatGPT reads live pages through browsing and connected shopping sources, and it also uses what it learned during training. The live layer is where you can actually move the needle, because it reflects your current pages and feed.
Why does my competitor show up in AI answers and I don't?
In most of our audits it's readable structured data and visible reviews. Competitors that rank in AI answers usually have complete Product schema, crawlable ratings, and descriptions that answer real buying questions. Stores missing those pieces get skipped.
Do I need to pay OpenAI or Perplexity to get recommended?
No. Organic recommendations come from your pages, feed, and reviews being readable and trustworthy. Paid placements are starting to appear in some shopping surfaces, but the free path is fixing your product data first.
How fast can changes show up in ChatGPT or Perplexity?
It depends on how often the assistants recrawl your pages. We usually see citation changes within a few weeks of fixing schema and descriptions. Fresh, accurate pages get picked up faster than stale ones.
If you run a Shopify or DTC store and want us to audit exactly why AI assistants are skipping your products, see how WRKNG Digital approaches agentic commerce here: https://wrkngdigital.com/agentic-commerce-landing-page.
