AI shopping starts with your product data: feeds, structured data and pages

What is AI shopping, and why does product data matter so much?

AI shopping describes search experiences in which a system compares products, summarizes options and shows prices and availability. Because the answer is assembled from data fields, the accuracy and consistency of your product data decide how your products are represented.

How are AI shopping answers assembled?

A system needs to know what a product is, who makes it, what it costs, whether it is in stock and how it differs from alternatives. It finds that information in structured fields, such as titles, brands, identifiers and prices, and in page text such as descriptions and reviews. When fields are missing or contradictory, the system may skip the product or describe it wrongly.

Why does this raise the stakes for feeds?

Search Engine Land argued that AI shopping starts with your product feed, not your product page, because feeds supply structured, regularly updated data to Google's shopping systems. The argument is a reminder that a beautiful page cannot compensate for a neglected feed.

What do the statistics say about AI in shopping results?

One vendor reported that AI Overviews appeared on about 14 percent of shopping queries, sharply up from late 2025. It is a single-source figure based on that vendor's tracked queries, so quote it only with attribution and check your own priority searches, which is the approach we recommend in our guide to AI Overviews and AI Mode.

How do you check your own exposure?

Search your top product and category queries in a private window, note whether AI answers or shopping features appear and record which merchants are shown. Repeat monthly.

Which businesses are affected?

Every store that sells online, and every local retailer that lists products. Even stores that do not run paid shopping ads can be eligible for free listings and merchant experiences when their product data is available to Google.

How does product data reach Google?

Product data reaches Google through your visible pages, through structured data on those pages and through feeds uploaded to Merchant Center. Google compares them, so consistency matters as much as completeness.

What does the page itself contribute?

The page shows shoppers the title, price, images, description, reviews and policies. Google reads that content, so anything important should be present as text in the HTML. Our guide to client-side versus server-side rendering explains why server-delivered content is safer.

What if the page changes after load?

If a script updates the price or availability after the page loads, make sure the markup and the visible value agree, and test what Google renders.

What does structured data contribute?

Structured data states the same facts in a machine-readable format, which can make a page eligible for product snippets and merchant listings. Google's ecommerce page, last updated December 10, 2025, recommends Product markup, ProductGroup for variants, Organization markup for business details, Review markup, BreadcrumbList for hierarchy and VideoObject for pages about a video.

Is markup required?

It is not required to have a page indexed, but it is recommended for participation in shopping experiences, and Google says it can help it understand the page.

What does the feed contribute?

A Merchant Center feed supplies product attributes in bulk and on a schedule, which suits large catalogs and frequent changes in price and stock. Google's page recommends setting up structured data for Merchant Center to improve participation in shopping experiences and suggests registering physical store locations and following the guidelines, such as sharing return policies on your site.

Which source wins if they disagree?

Do not rely on a winner. Mismatches can cause disapprovals or inaccurate listings, so remove the disagreement by feeding every route from one source of truth. Our article on feeds versus structured data compares the routes in more detail.

Which product fields matter most?

The core fields are a unique identifier, a clear title, a brand, a price with currency, availability, an image, a description and a product category. Missing or inconsistent values in these fields cause most data problems.

Which fields should every product have?

The table lists core fields and the practical advice for each. Check the current Google specifications for exact requirements, because they change.

Which field causes the most trouble?

Price and availability, because they change often and live in several systems. Automate the flow from one source, and monitor for mismatches.

Why do identifiers matter?

Global trade item numbers (GTINs) let systems recognize that your product is the same as one sold by others, so it can be grouped, compared and matched to reviews. Using the wrong or invented code can cause disapprovals, so use only codes from the manufacturer or an official registry.

What if a product has no GTIN?

Custom or handmade items may legitimately lack one. Follow Google's current guidance for such products and supply the other identifiers you do have, such as brand and part number.

How should you write titles?

A good title says what the product is, in the words shoppers use, with the attributes that distinguish it, such as size, color or material. Keep it factual and avoid capital letters, symbols and promotional phrases. Consistent title patterns across a catalog make products easier to compare.

Should titles match across the page and feed?

They should be the same or very close, so that systems recognize them as the same product. Differences invite confusion and can trigger review.

How should variants be represented?

Variants should be grouped under a parent product with attributes such as color and size, so that systems understand they are versions of one product. Google's guidance supports ProductGroup markup for this purpose, and feeds use group identifiers.

What does grouping achieve?

It prevents the same product from appearing as dozens of unrelated items, lets shoppers see the range and keeps reviews attached to the right parent. It also helps your own canonical strategy, which we discuss in our guide to product variants.

What must each variant carry?

Its own identifier, its price and availability, the attribute values that make it different and an image that shows it.

What mistakes occur with variants?

Mixing different products under one group, giving all variants the same identifier or showing a price that belongs to a different variant than the one selected are common. Each leads to wrong or rejected data.

How do you test variant data?

Pick products with many variants, inspect the markup for several selections and compare it with the visible page and the feed row.

How do shipping and return policies fit into product data?

Shipping and return terms are part of what shoppers judge, and Google supports markup for them. Google's Organization documentation, last updated September 8, 2026, describes default return and shipping policies at the organization level, with product-level overrides where terms differ.

What does Organization markup offer?

The Organization page says there are no required properties, that the markup belongs on the home page or a single page describing the organization and that you should use the most specific subtype, such as OnlineStore for an online shop. It supports hasMerchantReturnPolicy for default return terms and hasShippingService for default shipping terms.

When do you need product-level policies?

When specific products have different terms, for example oversized items or final-sale products, add the same properties in the product's merchant listing markup, as the documentation says.

Why must site policies match the markup?

Structured data must reflect what the site actually states and does. If markup promises free returns for 60 days while the policy page says 30, shoppers are misled and Google may distrust the data. Write the policy once, publish it clearly and generate the markup from it.

Where should policy pages live?

On clearly named, linked pages that shoppers and crawlers can reach, such as returns and shipping pages linked from the footer and product pages.

What about physical stores?

Google recommends LocalBusiness markup for physical stores, covering location and opening hours, and suggests registering store locations in Merchant Center. Keep hours and addresses consistent with your Business Profile.

How do reviews and trust signals feed AI shopping?

Reviews give systems and shoppers evidence of real experience. Use genuine, visible reviews marked up in line with Google's review snippet guidelines, and keep ratings consistent between your pages and your review platform.

What must reviews satisfy?

They must come from real customers, appear on the page and be marked up for the product they describe. Do not aggregate category-level ratings or mark up reviews of your own business about itself. Our guide to rating and price markup covers the rules.

What about reviews from other sites?

Syndicated reviews may be shown, but follow Google's guidelines on how they are marked up, and do not copy reviews you did not collect.

Which other signals help shoppers?

Clear contact details, transparent policies, secure checkout and consistent branding all reduce doubt. They do not replace accurate data, but they support the decision to buy.

Do questions and answers help?

Yes. Answered questions give detail that descriptions often miss, and they add natural language, as we explain in our article on reviews and Q&A as user-generated content.

How do you keep product data accurate at scale?

Create one source of truth, automate the flow to pages, markup and feeds, validate before publishing and monitor continuously. Manual copying between systems is the main cause of drift.

What does a good data pipeline look like?

Product information lives in a product information system or the ecommerce platform's database. Templates generate pages and structured data from it, and a scheduled export produces the feed. Validation rules check required fields, formats and consistency before anything is published. Changes to price and stock flow automatically.

Which validation rules help?

Required fields present, valid currency and availability values, titles within length guidelines, GTINs in valid format, images reachable and prices matching between the page and the feed.

How do you monitor Merchant Center and Search Console?

Review diagnostics in Merchant Center for disapprovals and warnings, and the product and merchant listing reports in Search Console for markup issues. Investigate sudden changes within days, because errors can affect thousands of products at once.

Who should own data quality?

Assign an owner who can coordinate merchandising, development and marketing, because problems usually cross team boundaries.

How do images and descriptions affect shopping listings?

Images and descriptions tell shoppers and systems what a product looks like and what it does. Accurate, original, clearly presented versions improve comprehension and reduce returns, while generic or copied text offers no advantage and may be matched to other sellers.

What makes a good product image?

A clean main image that shows the product itself, on a plain background, in high resolution and without overlays or watermarks is the safest choice for shopping surfaces. Extra images should show angles, details, scale and use. Name image files descriptively and write alt text that says what is in the picture, not a list of keywords.

Should variant images differ?

Yes. Each color or style should have its own image, so the selected option looks like the product the shopper will receive.

How should descriptions be written?

Start with what the product is and who it suits, then give the specifications, materials, dimensions and care. Use plain sentences and lists. Do not paste the manufacturer's text, which every seller shares, and do not include prices, promotions or shipping claims that will become stale.

How long should a description be?

Long enough to answer the questions shoppers ask, and no longer. Add detail where choices are complex, and keep simple products short.

What about attributes?

Structured attributes such as color, size, material and compatibility help filtering and comparison. Fill them consistently and use the same values across similar products, because "Navy", "Dark Blue" and "navy blue" split the same color into three.

Who should maintain attributes?

The merchandising or catalog team, with clear rules for allowed values, so that the data stays clean as products are added.

How should you handle stock changes and product lifecycle in the data?

Reflect availability accurately and promptly, keep pages live for products that will return and retire discontinued products cleanly. Wrong availability is a leading cause of poor shopper experience and disapprovals.

What should happen when a product sells out?

Mark it out of stock in the page, the markup and the feed at the same time, keep the page live and offer alternatives or a restock notice. Removing the page loses accumulated value, and leaving it marked in stock misleads shoppers.

How fast should updates flow?

As fast as your systems allow. For fast-moving stock, schedule frequent feed updates and make sure page markup changes with the page.

How do pre-orders and backorders fit?

Use the availability values that describe them accurately, state expected dates where you can and do not present a pre-order as in stock. Follow the current Merchant Center specification for the right values.

What about seasonal products?

Keep the same URL year after year, update the data and mark the product available only when it can be bought. Our category page guide explains handling for seasonal pages.

How do you retire discontinued products?

If a close replacement exists, redirect the old page to it permanently. Otherwise redirect to the most relevant category or return a 404 or 410 status, remove the item from feeds and sitemaps and update internal links. Google's crawl guidance says to return a 404 or 410 for permanently removed pages.

How do you prevent soft errors?

Do not show a normal-looking page with a "not found" message while returning a 200 status. Return a real error status for pages that no longer serve a purpose.

How do category and comparison content support AI shopping?

Category pages and comparison guides give systems and shoppers the context needed to choose between products, such as which features matter and who each option suits. Honest, structured guidance is more likely to be quoted than bare product lists.

What should a category page explain?

It should say what the product type is, how the main options differ, what to look for and how to measure or size. Use headings for each decision, tables for comparisons and short direct answers, as we describe in our article on category page content placement.

How honest should comparisons be?

Completely. Include trade-offs and say when a cheaper option is enough. Honest comparisons build trust and are less likely to be contradicted by other sources.

How do comparison tables help?

A table with consistent columns, such as capacity, power, warranty and price range, is easy to scan and easy to extract. Use real HTML tables and keep values consistent with product data.

Should you name competitors?

Comparing categories or features is usually enough. If you compare named products, be accurate and fair, and avoid claims you cannot support.

How do internal links connect it all?

Link categories to products, products to guides and guides back to categories using descriptive anchors. This helps shoppers and crawlers move through the buying journey, and it reinforces the topical structure of the site.

What testing and monitoring routine keeps data healthy?

Test each template before release, then monitor diagnostics weekly. A routine that takes an hour a week prevents problems that would otherwise surface as lost listings.

What should you test before a release?

Run a sample of products through the Rich Results Test, compare markup values with the page, check that the feed row for the same product matches, confirm that variants and sale items behave correctly and verify that returns and shipping markup match your policy pages. Record the results.

Which products belong in the sample?

Include a simple product, a variant product, a discounted item, an out-of-stock item and a product with no reviews. Edge cases reveal most template faults.

What should you monitor each week?

Merchant Center diagnostics for new disapprovals and warnings, Search Console reports for product snippets and merchant listings, and a sample of live search results for price and availability accuracy. Alert the team when error counts rise.

How do you respond to a spike in errors?

Identify the common cause, usually a recent release or a feed change, fix it at the source, resubmit and confirm that counts fall over the following days.

What does a product data clean-up look like in practice?

The example below is fictional and shows a method. The outcomes are invented for illustration.

What was the situation?

A fictional store selling kitchen equipment has Merchant Center warnings for price mismatches and missing identifiers, and its product pages show inconsistent titles compared with the feed. AI summaries of its products sometimes show outdated prices.

What did the audit find?

Prices in the feed update nightly while the site updates instantly, a plugin and the theme both output Product markup with different values and about a fifth of products lack GTINs although the manufacturer provides them.

What did the team change?

They moved to one pipeline that generates page content, markup and feed from the same data, increased the feed refresh frequency, disabled the duplicate markup, added manufacturer GTINs and standardized titles. They published return and shipping policies with matching Organization markup.

How did they follow up?

They watched diagnostics and enhancement reports clear over several weeks and tracked a sample of products in shopping results, noting that listings were more consistent. They did not claim a specific revenue uplift, because other factors changed at the same time.

How do multi-currency and international stores handle product data?

Each market needs its own correct price, currency, availability and policies, with the same product identity underneath. Mixing markets in one feed or one page creates mismatches that are hard to diagnose.

What changes by market?

Price and currency, shipping destinations and costs, return terms, language of titles and descriptions, tax display and sometimes availability and product variants. The identifier and brand usually stay constant, which helps systems recognize the same product in each market.

How should feeds be organized?

Use separate feeds or clearly separated rows per country and language, as the current Merchant Center specification allows, and make sure each page that a feed row points to shows the matching currency and language. Our guide to international structure on Shopify explains how URLs for each market can be organized.

What about structured data in each market?

Mark up the price and currency that a visitor to that market page will see. If a page shows euros, the markup should say euros, even when the base price is stored in dollars. Check each market's template separately.

How do you avoid cross-market mistakes?

Test sample products in every market, compare page, markup and feed values and include market checks in your release routine.

How do you handle taxes and duties?

Display prices as the local market expects, for example with tax included in some regions, and state clearly what is and is not included. Take legal and accounting advice for the markets you serve, because rules differ and change.

Who should own product data governance, and what should the policy say?

Product data crosses merchandising, development, marketing and customer service, so it needs one accountable owner and a short written policy. Without both, each team fixes its own part and the pieces drift apart.

What should the policy cover?

The policy should name the system of record for each field, the rules for titles and attributes, the allowed values for availability and condition, the schedule for feed updates, the validation checks and the process for fixing errors. It should also list the owner for each field and the escalation path when something breaks.

How long should it be?

One or two pages. A short document that people read beats a long one that nobody opens.

How do you handle changes from Google?

Google updates its Merchant Center specifications and structured data guidance from time to time, and Search Engine Land has reported on several such changes. Assign someone to check the official documentation monthly, note the last-updated dates, and update the policy and templates when requirements change.

How do you keep stakeholders informed?

Send a short monthly note covering error counts, notable changes and planned work. Visibility keeps data quality on the agenda.

What are the common mistakes with AI shopping data?

The common mistakes are mismatched prices, duplicate markup, invented identifiers, stale availability and neglected feeds. Each one confuses systems and shoppers.

Which mistakes create mismatches?

Separate systems for page, markup and feed, slow feed updates and prices changed by scripts after load are the typical causes.

Which mistakes create policy problems?

Using made-up GTINs, marking up policies that the site does not follow and displaying fake discounts breach guidelines and consumer rules.

How do you avoid them?

One source of truth, automated validation, honest data and regular monitoring.

What should you do in the first 30 days?

Start by measuring the gap between your pages, markup and feed, then fix the largest mismatches. A focused month produces visible improvement and a clear plan.

What is a practical sequence?

In week one, export a sample of 100 products and compare title, price, availability and identifier across page, markup and feed. In week two, fix the causes of the most common mismatches, usually in templates or feed rules. In week three, publish clear shipping and return policy pages and matching Organization markup. In week four, set up weekly diagnostics monitoring and record the baseline for your tracked queries. Repeat the sample check each month.

What results should you expect?

Fewer diagnostics warnings, more consistent listings and a documented process. Revenue effects depend on many factors and should be judged over months.

How does AH SEO Agency help with product data for AI shopping?

We audit how product data flows from your platform to pages, markup and feeds, remove mismatches, set validation rules and monitor diagnostics. We then connect the work to category and product page SEO so that your data and your content support each other.

Where can you learn more?

See our product page SEO service, read our GEO versus SEO guide or request a free SEO audit.

Summary: what should stores do to prepare for AI shopping?

Make product data accurate, complete and identical across pages, structured data and feeds. Use Google's recommended markup, publish clear policies and monitor diagnostics, and treat vendor statistics as directional.

• AI shopping answers are assembled from product data fields.
• Pages, markup and feeds must agree, so generate them from one source.
• Use Product, ProductGroup, Organization, Review and BreadcrumbList markup as Google recommends.
• Core fields are identifiers, titles, brand, price, availability, images and category.
• Monitor Merchant Center and Search Console, and track your own products in results.