GEO for online shops is the work on product information that AI assistants such as ChatGPT, Gemini and Perplexity can find, read and compare when a customer asks which product suits them. It takes full specifications, honest limitations, category explanations and buying guides, plus feeds and structured data that match the page. Whether an assistant mentions your product is up to that system.
Which product questions do customers ask AI assistants?
A customer using an AI assistant rarely types just a product name. They describe their situation. Say a customer wants a rain jacket and asks: “Which rain jacket keeps me dry cycling 12 kilometres to work and fits over a blazer?” That question contains requirements (waterproof, suitable for cycling, a generous cut), a use and a limit.
Which rain jacket keeps me dry cycling 12 kilometres to work and fits over a blazer?
Look for a long cycling rain jacket with an extended back and room in the shoulders. Regenkleding De Bui says of its Forens model that size L fits with a blazer underneath, that it has a waterproof rating and underarm vents, and that it is not suitable for running. If you cycle in the dark, check for reflective details too.
Bronnen: debui.example.com/forens-cycling-rain-jacket
The questions fall roughly into five groups:
| Type of question | Example | What the assistant needs |
|---|---|---|
| Suitability | “Is this vacuum cleaner good for dog hair on carpet?” | Intended use, test situation, limitations |
| Comparison | “What is the difference between merino and synthetic base layers for hiking?” | Material properties, pros and cons |
| Budget | “Best coffee grinder for filter coffee on a set budget” | Current price, key specifications |
| Compatibility | “Does this battery fit my type X drill?” | Exact model numbers, connections |
| After purchase | “How long is the return period at this shop?” | Shipping and return conditions |
Almost every one of these questions is about information that belongs on a good product page. GEO for an online shop therefore starts with your own pages.
Where do AI assistants get product information?
Each provider works differently, and not everything is documented. What the providers describe themselves:
Google. For AI Overviews and AI Mode, Google says in AI features and your website that the usual SEO basics still apply and that you need no special schema or separate file. Among its good practices, Google lists structured data that matches the visible text and up-to-date data in Merchant Center.
OpenAI. According to the overview of OpenAI crawlers, ChatGPT uses OAI-SearchBot to show websites in its search features. Sites that block OAI-SearchBot do not appear in ChatGPT’s search answers, although they can still show up as a navigational link. OpenAI also publishes a product feed specification with required fields such as title, description, price and availability, and optional fields such as GTIN, reviews and returns policy.
Other assistants. For Perplexity, Copilot and Claude: read their own documentation and do not count on a method they do not describe themselves.
The common thread: assistants work with what they can retrieve from your site or from a feed. If information is missing, sits only in an image, or the page and feed contradict each other, the system has less to work with.
How do you make product information complete?
A product page that an AI assistant can use answers a customer’s questions without them having to call. Go through these parts for each product:
| Part | Why it matters | Example |
|---|---|---|
| Your own description | A manufacturer’s text often appears at other shops too; your explanation does not | What you recommend this product for |
| Specifications as text | Numbers in an image or PDF are hard to read | Weight, dimensions, material, power |
| Intended use | Links the product to a situation | “For daily commuting up to 20 km” |
| Limitations | Answers “is this suitable for...” honestly | “Not dishwasher safe”, “runs small” |
| Size chart | Answers a standard question for clothing | Sizes in centimetres, per fit |
| Compatibility | Needed for parts and accessories | Exact model numbers |
| Identifiers | Links your product to the same product elsewhere | Brand, GTIN, model number |
| Price, stock, delivery time | Current and identical on the page, in the schema and in the feed | |
| Shipping and returns | Often the last question before buying | Costs, period, conditions |
| Reviews | Buyers’ experiences | Visible on the page, not only in a widget |
The full list is in the product page checklist.
Why do honest limitations help?
Many shops only write about benefits. A customer who asks an AI assistant whether a product is suitable wants to know exactly where it stops. If your page says “suitable down to minus 10 degrees” and another shop says nothing, the assistant has a concrete answer from you.
In its guidance on helpful, reliable content, Google asks, among other things, whether content shows that someone has actually used the product. Limitations from your own experience, reasons for returns and questions from customers are exactly that information. Use them.
A product description without and with limitations
VoorThis stylish travel mug keeps your drinks hot for hours. Made from premium materials. Perfect on the go!
NaDouble-walled stainless steel travel mug, 350 ml. Keeps coffee hot for a morning with the lid closed; with the lid open it cools faster. Leakproof in a bag when the flap is locked. Not dishwasher safe and not for fizzy drinks. Fits most car cup holders, but not narrow bicycle holders.
What should you do with reviews?
Reviews provide experiences you cannot write yourself: how the fit turns out, how long something lasts, how the delivery went. Three points:
- Put reviews on the product page as text. Reviews that only load through JavaScript are not in the HTML the server returns.
- Link reviews to the right product or the right variant.
- Only include ratings in your structured data when they are visible on the page.
How do category explanations and buying guides help?
Product questions often start broader than a single product. “Which walking boot for the Alps?” is first about a type of boot, and only then about a model. Two kinds of page come into play:
- Category pages with an explanation. A short text that says which types exist and how to choose, with links to the products. See category pages that get found and help people choose.
- Buying guides and comparisons. A page that explains honestly which product suits which situation, including when a cheaper model is enough.
What should you do with product feeds?
A product feed is a file containing all your products and their details, which you submit to a platform.
Google Merchant Center. In Share your product data with Google, Google describes two routes: structured data on your pages and a feed in Merchant Center. According to the product documentation, combining both increases the chance of appearing in more search experiences and helps Google verify your data. Google warns that the feed and the site can drift apart, for example when a sold-out product is still in stock in the feed, and offers a setting that updates the feed automatically from your site.
OpenAI. OpenAI’s product feed specification exists, but the overview does not state who can sign up. Check OpenAI’s current terms before you put time into it.
Apply the same principle to every feed: price, stock and conditions identical to the page.
One product, three places where the data has to match
| Product page | Structured data | Feed | |
|---|---|---|---|
| Name | Travel mug Onderweg 350 ml, moss green | name with the same text | title with the same text |
| Stock | In stock | availability: InStock | availability: in_stock |
| Variant | Colour moss green, own URL | Variant with its own url and color | Separate item with colour and link |
| Identification | Item number and EAN visible | sku and gtin13 | gtin |
| Returns | Returns conditions below the button | Return policy in the Offer or at organisation level | Returns data as the specification describes |
Which technical points have to be right?
- Let crawlers in. Check in your robots.txt that search crawlers such as Googlebot, Bingbot and OAI-SearchBot are allowed to fetch your product pages. The difference between search crawlers and training crawlers is explained in allowing or blocking AI crawlers with robots.txt.
- Information in the HTML. Put the description, specifications and price in the HTML the server returns, not only in tabs that load after JavaScript. More on this in Can AI crawlers read your website?
- One version of the structured data. Two JSON-LD blocks with different data for the same product make it unclear which value applies. See which structured data does an online shop need?
What do Shopify and WooCommerce handle?
Shopify. Shopify has a Liquid filter called 0 that outputs product data as JSON-LD, as ProductGroup for products with variants. Whether your theme uses that filter, and whether apps add a second block next to it, differs per shop. Check it.
WooCommerce. The WooCommerce core generates basic markup automatically. For variable products, according to the WooCommerce documentation, that is one combined block. Plugins for SEO or feeds often add their own markup.
On both platforms, you write the descriptions, specifications and size charts yourself. The platform provides the structure, not the content.
How do you measure whether it works?
You cannot see how often an AI assistant mentions your product without checking. A useful check records which questions you ask, in which assistant, with which settings and when, and whether your shop is mentioned or linked as a source. Repeat it with the same questions. You can also follow traffic from AI assistants in your analytics. How that works is explained in how do you measure your visibility in AI answers?
How Outseen approaches this
Outseen has a standard approach for online shops on Shopify and WooCommerce. Atlas, Outseen’s own research engine, examines the public code of fetched pages, including robots rules, indexing instructions, canonicals and supported JSON-LD such as product and breadcrumb, and records for each finding what was found on which URL and when. Anything that only appears after JavaScript falls outside it. We check AI answers by hand, on agreed questions, environment and measurement moment.
What does this mean for your online shop?
Pick ten products that sell well or generate a lot of questions and review those first. You do not need to tackle your whole range at once; a small set quickly shows where the gaps are, and in the rest of your products they are usually in the same places.
How to check one product page
- The description is in your own words and says who the product is for.
- Specifications are on the page as text, not only in an image or PDF.
- At least one honest limitation is included.
- Every variant has its own URL with the right photo, price and stock.
- Price and stock match on the page, in the structured data and in your feed.
- Reviews are in the HTML as text.
- Delivery and returns conditions can be reached from the product page.
- Your robots.txt allows Googlebot, Bingbot and OAI-SearchBot.
Then put a few real customer questions about those products to the AI assistants your customers use and write down what they answer. Repeat that with the same questions after your changes. A longer list for reviewing your whole site is the GEO checklist.