AI Engines · 9 min read · July 15, 2026
Product data for AI: how your items get cited by ChatGPT and Perplexity
When a shopper asks ChatGPT "Which running shoes for overpronation under 120 euros?", your Google ranking has nothing to do with the answer — what matters is whether the AI can actually parse your product data. Generative Engine Optimization makes your catalog readable for language models: structured markup, concrete attributes, verifiable facts. Get this wrong and you disappear before the shopper ever reaches a click.
Why product search is shifting right now
Your customers search differently than they did just two years ago. Instead of typing 'best coffee machine with grinder' into Google and clicking through ten test reports, they ask ChatGPT or Perplexity directly: 'Which fully automatic espresso machine under 500 euros is quiet enough for an office?' The AI answers with three concrete models, and only the brands that make that answer get noticed at all. The classic ten-blue-links moment disappears, and with it the chance of still getting picked from position four or five.
For online shops, this is a quiet threat. Your Google Analytics numbers may still look stable, but the question is for how long. Perplexity cites its sources visibly, and ChatGPT increasingly surfaces product mentions through its search mode. If a competitor is named there consistently and you aren't, you lose revenue at a point that shows up in no classic SEO dashboard. GEO isn't a trend you can wait out — it's the channel where market share is being redistributed right now. Search behavior is already moving in this direction: more than two-thirds of Google searches now end without a click to any website, up sharply from a couple of years ago, and results with an AI summary attached see a much lower click-through rate for the top organic listing than results without one.
The good news: the lever sits exactly where you already have homework to do as a shop — your product data. Language models reward clarity, completeness, and structure. Whoever describes their items cleanly wins twice: with the AI, and with the human who ends up buying anyway.
How ChatGPT and Perplexity actually read your product pages
A A language model doesn't see your shop the way a customer does, with images, hover effects, and discount badges. It reads text and structured data. Whatever sits in a JavaScript carousel, inside a graphic, or buried in a PDF spec sheet effectively doesn't exist for the AI. If your strongest selling point — say, 'waterproof to 10 bar' — only appears as an icon with no alt text on the product image, it won't show up in any answer. The rule of thumb: what isn't present as clean text won't get cited.
What actually decides this is structured markup, specifically Schema.org Product, with fields like name, brand, description, offers, price, availability, and aggregateRating. This is the format machines have used to exchange product data for years, and AI crawlers rely on exactly that. A shop that ships valid Product markup on every item makes the AI's job easy: price, availability, and rating are unambiguous, instead of having to be guessed out of vague running text.
Don't judge this from the gut. Take a real product page, strip it down to plain text — your browser's reader view works fine — and ask yourself: are all the facts here that a customer needs to decide? If you have to guess whether the shoe runs true to size or narrow, the AI can't know either.
Your customers' questions are your content plan
People ask AI in full sentences, with context. Not 'women's winter jacket', but 'warm winter jacket for cycling at minus 10 that doesn't make me look like the Michelin man'. That kind of long query is gold for you — it reveals the occasion, the constraint, and the emotional criterion all at once. A shop whose product copy speaks to that situation ('breathable enough for the bike commute') matches the question far more precisely than a description that only lists material and size.
Collect these questions systematically. Your support tickets, the search-field logs in your shop, the reviews, and the return reasons are a goldmine. If dozens of customers ask whether the espresso machine works with ground coffee instead of only pods, that answer belongs verbatim in the product description and in an FAQ on the page. It's exactly this phrasing that the AI picks up, because it's linguistically closest to the customer's question.
Think in jobs, not categories. A customer doesn't buy a 'Bluetooth speaker' — they want 'music in the bathroom without worrying about splashes'. Write your product copy around these jobs and you get recommended for exactly the situational questions that make up most purchase advice in the AI era.
Attributes that matter: specifics beat marketing copy
Language models love hard, comparable facts and struggle with superlatives. 'Best sound in its class' is worthless to an AI, because it can't be verified. 'Battery life 18 hours, weight 240 grams, IPX7 waterproof', on the other hand, is exactly what lands in a comparison answer. If a customer asks 'Which headphones last a full workday without charging?', the AI can only name your item if that number is machine-readable somewhere.
So build out your attributes completely and consistently. For fashion, that means fit, material, care instructions, sustainability certification, and real size charts instead of just S to XL. For electronics, it's technical specs, compatibility, and what's in the box. Gaps are costly: if one of twenty similar products is missing the compatibility detail, that's the one that drops out of every 'Does this fit an iPhone 15?' answer, even if it technically does.
An often-overlooked field is an honest comparison within your own assortment. A short paragraph like 'Model A is lighter, Model B has the stronger battery' helps the AI place your product in the right recommendation, and positions you as a source that thinks alongside the customer instead of just selling to them.
Reviews and real usage data as a trust signal
AI systems weight signals that point to real, lived experience. Authentic customer reviews are ideal for this, because they contain language no marketing team writes: 'runs half a size small', 'stretched out a bit after three washes', 'great for wide feet'. That exact phrasing answers later customer questions and turns your product page into a source the AI can rely on. A shop that integrates reviews in a structured, marked-up way hands the AI that proof directly.
What matters is honesty over polish. A wall of five-star reviews with not a single critical note reads as implausible to humans and models alike. An honest four-star average with a legible weakness ('build quality is excellent, but the manual is thin') builds more trust, and gives the AI context for who the product suits and who it doesn't.
Add concrete usage scenarios from real life on top of that. A short practical note — 'tested over two weeks as a commuter backpack' or 'often bought by customers for a home office setup' — carries evidentiary weight that manufacturer specs alone never reach.
The most common mistake: thin, interchangeable product copy
Many shops copy the manufacturer's description word for word. The problem: the same sentences then show up at a hundred other retailers. The AI has no reason to cite you specifically if your text is identical to your competitor's. Duplicate content was already a problem in classic SEO. In generative search it's even more costly, because the model simply favors the source with the most original contribution and context.
The way out isn't more text at any cost — it's your own added value. Layer your perspective on top of the manufacturer data: who the product is worth it for, who it isn't, what it pairs with, and the typical mistakes people make when buying it. For a tent, for example: 'great for festivals, not storm-stable enough for alpine trips'. No manufacturer copy delivers that kind of judgment call, and it's exactly what makes you citable.
Focus on substance over keyword stuffing. Sentences that cram in 'cheap women's winter jacket buy' three times do actual harm — language models recognize unnatural patterns and downgrade the source. Write for the human asking the question, and it will also be right for the machine.
Getting into the answer: feeds, FAQs, and freshness
Beyond your product pages, three places are worth sharpening up. First, a clean, current product data feed — many AI systems pull from merchant feeds and marketplace data, and outdated prices or wrong availability there will lead the AI to wrongly write you off as 'not available'. Second, structured FAQ blocks per product that answer real customer questions verbatim and are marked up with FAQPage schema.
Third, freshness matters more than it used to. A guide titled 'The best gas grills of 2024' won't get recommended in 2026 if those models are off the market. Keep the year current, remove discontinued items, and refresh your test-winner references. AI systems favor visibly maintained sources, because outdated recommendations are a trust problem for them too. A 'last updated' date is more than cosmetic.
Measure this concretely. Regularly ask ChatGPT, Perplexity, and Gemini the ten most important purchase questions in your category and note whether, and how, you're named. This manual spot check replaces no tool, but it shows you faster than any dashboard whether your product data is actually reaching the AI.
A concrete roadmap for the next few weeks
Start small and measurable. Take your twenty highest-revenue products and work through them completely: valid Product markup, complete attributes, a genuine added-value paragraph, three real FAQs, and integrated reviews. In practice, these twenty pages tend to drive the bulk of your revenue, and they're the fastest way to start appearing in AI answers without touching your whole catalog at once.
After that, systematize it. Define an attribute template per category so nothing gets forgotten, and build the upkeep into your product-creation process — new items shouldn't go live without complete data in the first place. In parallel, set up a fixed monthly check with the AI systems to track progress and catch regressions.
Stay honest with yourself: GEO doesn't replace a good assortment or a fair price. It makes sure a good product actually gets found. The effort pays off twice, because the same clear, honest product data also convinces your human customers and lowers your return rate.
Common questions
Is it enough to just add Schema.org markup, or do I also need to rewrite the text?
Markup alone isn't enough. It helps the AI read price, availability, and rating reliably, but the actual recommendation comes from the content. If your description is thin or identical to a competitor's, even the cleanest markup won't do much. You need both: technically clean structured data and content that's independently written, concrete, and full of real attributes and usage scenarios.
My product copy comes from the manufacturer. Is that a problem for AI visibility?
Yes, if you use it unchanged. The same sentences then show up at many other retailers, and the AI has no reason to cite you specifically. You don't need to rewrite everything — just add your own perspective: who the product suits, what it pairs with, common buying mistakes. That original contribution is what makes you the preferred source.
How do I even tell whether ChatGPT or Perplexity recommends my shop?
The fastest way is a manual spot check. Phrase the ten most important purchase questions in your category the way real customers would ask them, and put them regularly to ChatGPT, Perplexity, and Gemini. Note whether you're named or linked, and which competitors show up. Perplexity shows its sources directly, which makes this easier to judge. It's no substitute for a tracking tool, but it gives you an honest read on your current AI visibility.
Read on