Ask ChatGPT for "the best running shoes for flat feet under $150" and you don't get ten blue links — you get a shortlist of named products, often with prices, images, and a sentence explaining why each one fits. The same is happening in Perplexity, Google's AI Overviews, and Copilot. The stores in those answers didn't get there by luck: their product data was structured, consistent across the web, and open to AI crawlers. Here's how to make your catalog one of them.
Why do AI answers name specific products now?
Because "best X for Y" questions are exactly what answer engines are built to resolve. A shopper asking "best ergonomic office chair for a short person" doesn't want a category page — they want three candidates and a reason to pick one. Engines build that shortlist from what they can verify: your product pages, structured data, merchant feeds, marketplace listings, and third-party reviews.
That verification step is the whole game. An engine won't confidently recommend a product it only half-understands. If your product exists as marketing copy buried in JavaScript, with no schema, no feed, and a hidden price, the engine skips it and recommends a competitor it can actually describe. (For the mechanics, see how AI answer engines work.)
What schema markup do product pages need?
Four types, nested together: Product, Offer, AggregateRating, and Review. Product schema is machine-readable labeling from schema.org that tells crawlers, unambiguously, what the item is: name, brand, SKU, GTIN if you have one, image, and a real description. The nested Offer carries price, currency, and availability. AggregateRating and Review carry your rating count and score — the social proof engines lean on when deciding which of five similar products to name.
Two rules keep this markup working for you instead of against you. First, it must match the visible page — a schema price that disagrees with the displayed price reads as unreliable data. Second, it must stay current: availability set to InStock on a discontinued item is worse than no markup at all. Most platforms (Shopify included) emit baseline Product schema automatically, but rating, review, and GTIN fields are often missing until you add them. Our schema markup guide covers implementation and testing.
How should product descriptions change for AI?
Lead with the answer. The first sentence of every product description should state what the product is, who it's for, and the one thing that differentiates it — because that sentence is what an engine lifts into its answer. "The Meridian 2 is a stability running shoe built for flat-footed runners who overpronate, with a firmer medial post than most shoes in its price range" is quotable. "Experience the pinnacle of performance" is not.
Two habits matter most:
- Write original copy. If you paste the manufacturer's description that forty other retailers also use, an engine has no reason to cite you over any of them — or over the manufacturer.
- Write spec sentences, not spec tables alone. "Weighs 260 grams in a men's size 9" and "fits desks 24 to 30 inches deep" are the factual, self-contained statements engines extract. Prose carrying hard numbers is the most quotable format there is — more on this in writing content AI can quote.
What comparison content should a store publish?
The content that mirrors how shoppers actually ask. Three formats consistently earn citations:
- "Best X for Y" buying guides — one page per real buyer situation (best standing desk for tall users, best beginner espresso machine under $300), with named picks and stated reasons.
- Head-to-head comparisons — "Product A vs. Product B" pages that answer honestly, including when the competitor wins on some criterion. Honesty is what makes the page credible enough to cite.
- Fit and compatibility answers — "does X work with Y," sizing questions, material questions. Low glamour, high intent, and rarely answered well anywhere.
You're not just optimizing product pages; you're publishing the editorial layer engines quote when they explain why a product is the right pick.
Do feeds and marketplaces actually help?
Yes — as corroboration. Answer engines trust product facts they can confirm in more than one place. A Google Merchant Center feed, a Bing merchant feed (which matters for Copilot), marketplace listings, and review platforms all repeat your product's name, price, and availability. When those sources agree with your site, the engine's confidence in your data rises; when they contradict each other — old prices in a stale feed, a renamed product still under its old title on a marketplace — confidence drops.
Treat feeds as a first-class output, not an afterthought: same titles, same GTINs, same prices as the site, refreshed automatically. And keep your own product pages canonical and strong — if the richest data about your product lives on a marketplace, the engine may recommend the product but cite the marketplace, and the margin goes with it.
Why does price transparency matter so much?
Because engines want to state a price, and they favor products whose price they can state. "Contact us for pricing," login-gated prices, and prices rendered only by client-side scripts all push your product off the shortlist in favor of one the engine can describe completely. Publish the price on the page, mirror it exactly in the Offer schema and your feeds, and update all three together when it changes.
Are blocked crawlers killing your product citations?
Quite possibly — it's the most common self-inflicted wound we find in e-commerce audits. In the scramble over AI training data, many stores blocked crawlers like GPTBot, ClaudeBot, and PerplexityBot, sometimes via a CDN toggle nobody remembers flipping. But crawlers also power retrieval: the live lookups engines make when composing shopping answers. Block them and your products can't be read, verified, or recommended — the engine simply builds its shortlist from stores that let it in.
Check your robots.txt and your CDN's bot-management settings, and decide bot by bot what you're comfortable allowing. Our guide to AI crawlers and robots.txt walks through each crawler and what it feeds.
What are the early AI shopping surfaces?
ChatGPT shows product cards with images and prices for shopping queries, Perplexity has a shopping experience with product results, and Google is threading its Shopping Graph through AI Overviews and AI Mode. These surfaces draw on the same inputs covered above — structured data, feeds, open crawler access, corroborating reviews — and they're young enough that most catalogs are barely optimized for them. Stores that get their data in order now compete against thin fields; the ones that wait will fight incumbents.
Frequently asked questions
Do I have to pay to appear in AI shopping answers?
Not currently. Today's product recommendations in ChatGPT, Perplexity, and AI Overviews are organic — drawn from crawled pages, feeds, and reviews rather than paid placement. Ad formats may come, but right now visibility is earned through data quality.
My products are already on Amazon. Isn't that enough?
No. Marketplace listings corroborate your data, but if the marketplace is the best source about your product, the engine cites the marketplace and the sale lands there. Your own product pages need to be the richest, most current source.
How do I know whether AI engines recommend my products?
Ask them. Run your top twenty "best X for Y" questions through ChatGPT, Perplexity, and Google's AI results monthly, and note whether you're named, a competitor is, or nobody is. Watch your analytics for referrals from AI domains as a second signal.
Will AI answers just steal my traffic?
Shopping queries behave differently from informational ones. A shopper whose AI answer names your product arrives pre-sold — fewer clicks overall, but the ones you get carry high intent. The real risk isn't AI answers existing; it's your competitor being the name in them.
Find out if AI engines can even see your catalog
Most stores have at least one silent blocker — a bot rule, broken Product schema, a stale feed — standing between their products and AI answers. Our free AEO audit checks your crawler access, structured data, and answer visibility and tells you exactly what's in the way. If you want the fixes handled, remediation is $99/hour and ongoing monitoring is $19/month — but the audit itself costs nothing and takes minutes to request.