Short answer: to get recommended in ChatGPT Shopping, expose the signals the agent reads before it picks a product. Put complete Product and Offer schema (price, currency, availability) in the server-rendered HTML, add a product identifier like a GTIN, mark up reviews with AggregateRating, state returns and shipping clearly, and let AI crawlers read the page in robots.txt. No tool can guarantee a recommendation, but a page missing these rarely gets picked.
AI shopping is not a future thing you can wait on. ChatGPT Shopping is live to shoppers in the United States, Perplexity recommends products inside its answers, and Google AI Mode surfaces product comparisons above the classic results. Across all of them the core question the agent is answering is the same: which product is safe, relevant, and complete enough to put in front of a buyer? Your job is to make the answer obviously yes.
The good news for a smaller store is that this is a data game, not an advertising budget game. An agent cannot recommend what it cannot read, and it will not recommend what it cannot verify. If your data is clean, current, and machine-readable, you compete on merit with much larger brands whose pages are often a mess of client-side scripts. Let us walk through the five things that decide it.
Give the agent complete Product and Offer data
Product schema is how an AI shopping agent reads what your item is and what it costs. At a minimum, your Product JSON-LD needs an Offer with price, priceCurrency, and availability. Without a readable price, your item gets skipped, because an agent will not recommend something it cannot quote. Without availability, it treats your listing as risky and moves on.
Most themes output some Product schema, but two gaps are common: the price is missing or stale, and the availability field is absent. Fill both, and keep availability synced with real stock so it never goes stale. Stale pricing and wrong availability are two of the fastest ways to get dropped from an AI recommendation.
Verify a live product page with Google's Rich Results Test. If it reports a valid Product with an Offer, an AI crawler can read the same data.
Add a product identifier so AI can match you
A GTIN, MPN, or SKU is the single best matching signal you can add. It lets an agent connect your listing to the same product across the web and merge it into its product graph. A product with no identifier floats unmatched, which makes an agent far less confident recommending it over a rival that carries a clean barcode.
If you sell branded goods, you almost certainly have the GTIN or UPC already. Put it in the Product schema. For your own products, a consistent SKU still helps the agent treat the listing as a real, catalogued item rather than a loose page.
Expose reviews as structured data
Reviews are one of the strongest recommendation signals there is. Agents lean toward products with visible, structured ratings because that is the social proof they cite when they choose you over a competitor. The important word is structured. A review widget that looks nice but outputs no schema does little for AI shopping.
Use a reviews app that renders visible stars and writes AggregateRating and reviewCount into the page. Authentic reviews that mention specific use cases, and both pros and cons, help an agent understand whether the product actually delivers, which is exactly what it is trying to figure out before it recommends you.
State your trust signals in plain, readable terms
Before an agent recommends a merchant, it checks that you look legitimate. Make the basics easy to find and, where possible, machine-readable: a clear return policy, shipping and delivery terms, your company name and contact details, and an accurate in-stock signal. Add Organization schema so the agent recognizes you as a real brand entity, and add return and shipping details to the Offer where your platform supports it.
These are not just AI signals. They are the same trust cues a human buyer looks for, so tightening them helps conversion on your normal traffic too.
Let AI crawlers actually read the page
None of the above matters if the crawler cannot reach your content. Two things quietly block it. First, robots.txt: if you disallow GPTBot, OAI-SearchBot, ClaudeBot, or PerplexityBot, that engine literally cannot read your products, so it can never recommend them. Check your rules and open them for the shopping crawlers you want.
Second, and more common, is JavaScript rendering. Many AI crawlers do not run JavaScript. If your Product schema is injected in the browser after the page loads, the crawler sees an empty page even though a human sees a full one. The fix is to render Product and Offer JSON-LD in the initial HTML response, server-side.
This is the gap most stores miss, because the page looks perfect to them. It is only invisible to a crawler that skips JavaScript, which is exactly the audience you are trying to reach.
Check your product page free, in seconds
Our free AI Shopping Readiness Checker reads a product URL and grades it on exactly these signals: Product and Offer schema, price and availability, a GTIN, reviews, trust signals, and whether AI crawlers can read the page. It even detects if your schema is client-rendered and invisible to a non-JavaScript crawler. No signup, no AI tokens, nothing stored.
Check my product free →Once you know where your product page stands, the fixes cluster into a few tools. Missing reviews or social proof? PopBoost adds visible, schema-backed rating widgets. No offers or bundles with clean pricing? EZBundle. And the one signal agents drop you for fastest is stale availability, which is really an inventory problem. EZStock keeps stock and availability accurate so your in-stock signal stays true for both shoppers and AI.
If you want the deeper mechanics, the companion guide on product schema for AI shopping shows the exact JSON-LD to paste, and the AI shopping readiness checklist turns all of this into a page-by-page pass. For the bigger picture of why this is happening, see what agentic commerce means for a Shopify store.
Frequently asked questions
How do I get my products recommended in ChatGPT Shopping? Expose the signals it reads: complete Product and Offer schema with price and availability, a product identifier, reviews marked up with AggregateRating, clear returns and shipping, and robots.txt that lets the AI crawlers in. Then keep price and stock current. No tool can guarantee a recommendation, but a page missing these signals rarely earns one.
Does ChatGPT read JavaScript-rendered product pages? Often no. Many AI shopping crawlers do not run JavaScript, so schema injected in the browser can be invisible to them. Render Product and Offer JSON-LD in the server response. You can check whether yours is server-rendered with the free checker above.
How often should I update my product data? Keep price and availability current at least daily. Products with stale pricing or wrong availability get dropped from AI recommendations, because an agent will not risk sending a buyer to a listing it cannot trust.