AEO for E-commerce

Why AI Citations Matter for Online Stores

Product research has always been the part of shopping that happens before the shopping. Buyers read comparisons, check dimensions, look for the catch in the returns policy, and try to work out whether the cheaper option is actually worse. That research increasingly happens inside an assistant rather than across a dozen browser tabs. Someone asks ChatGPT, Gemini, Copilot or Perplexity for the best option in a category, and the assistant answers in prose, names a handful of products, and links to the sources it drew on.

The commercial consequence is simple. If your product pages are not among the sources an assistant can read and quote, you are not in the shortlist, and you never get the chance to compete on price, photography or checkout experience. The shopper who arrives from an assistant has usually already done the comparison step, which is exactly the traffic every merchant wants. Being cited is also durable in a way that paid placement is not: the same page keeps earning mentions as long as it remains the clearest available answer.

The uncomfortable part is that most ecommerce catalogues are structurally unsuited to this. A typical product page is a gallery, a price, a variant selector, a short marketing paragraph written by a copywriter, and a wall of user reviews loaded by JavaScript. There is almost no extractable text. An assistant asked whether the item fits a particular use case has nothing to work with, so it quotes a review site, a marketplace listing, or a competitor who bothered to write things down. Thin pages are not penalised; they are simply invisible.

The Shopping Queries Worth Winning

Assistants get asked a fairly predictable set of commercial questions. Each one maps to a page type you can deliberately build.

• Best X for Y. Best running shoes for flat feet, best espresso machine for a small kitchen, best laptop bag for a 16-inch machine. These are constraint queries, and they are won by pages that state the constraint explicitly.
• X vs Y comparisons. Two named products, or your product against a well-known competitor. Assistants love a page that lines up both options attribute by attribute.
• Is X worth it. A value judgement. The answer requires honest discussion of trade-offs, not marketing copy.
• Sizing and compatibility. Does this fit a 2019 model, what size should someone between two sizes order, will this cable work with that port. Enormously high intent and usually unanswered on-site.
• Materials and specifications. What is it actually made of, what does it weigh, what is the capacity, what is the thread count, what is the power draw.
• Shipping, returns and warranty. How long is delivery, who pays return postage, how long is the guarantee, is the packaging recyclable. Assistants are asked these constantly and answer from whatever policy page they can parse.
• Budget-constrained recommendations. Under a hundred pounds, cheapest option that still does the job, best value rather than best overall.

Notice that only two of these are satisfied by a conventional product page. The rest need content that most stores have never written.

A Content Strategy That Earns Citations

Make product pages genuinely readable

Every product should carry a specification table in real HTML, not an image of a size chart and not a PDF. Dimensions, weight, materials, capacity, power, compatibility, country of origin, care requirements. Use a <table> with clear row labels so that an assistant can lift a single attribute without guessing. Follow it with two or three paragraphs of prose that answer the obvious questions a shop assistant would field: who this suits, who it does not suit, and what people most often get wrong when ordering.

Critically, the text must be in the initial HTML response. Content that only appears after a client-side fetch, inside an accordion that renders on click, or in a third-party review widget is frequently missed. If your reviews live in a widget, mirror the key content server-side.

Build the layer above the product

Category explainers and buying guides are what assistants quote for the best X for Y queries. A good guide names the decision criteria, explains what each one means in practice, and then recommends specific products against specific constraints. Sizing and fit guides deserve their own pages with measurement instructions and a between-sizes recommendation. Care, maintenance and compatibility pages capture a long tail of post-purchase questions that quietly build topical authority.

Write honest comparisons

Comparison pages that only flatter your own product read as marketing and get treated accordingly. Pages that concede where a competitor genuinely wins are far more likely to be quoted, because they are more useful. Say plainly that the rival is lighter, or cheaper, or better supported in a particular region, and then explain the cases where yours is the better buy. Keep the comparison factual and current.

Publish substantive review content

Aggregate star ratings alone give an assistant nothing to say. Surface the text of real customer reviews as crawlable HTML, and where you can, add editorial context: common themes, recurring complaints, how the product held up over time. Long-form owner feedback is one of the most citable assets an ecommerce site can have.

Treat policy pages as content

Shipping, returns and warranty pages should be written in specific, quotable sentences. State the delivery windows, the return period, who pays for return shipping, the condition requirements, and the warranty term. Vague policy language produces vague answers or, worse, an assistant defaulting to a marketplace listing of your product instead of your own store.

Schema Markup for Ecommerce

Structured data does not create authority on its own, but it removes ambiguity about what a page describes. The types that matter most for retail:

• Product with a nested Offer carrying price, priceCurrency, availability, itemCondition and priceValidUntil. Add gtin, mpn, sku and brand wherever you have them so the item can be matched across sources.
• AggregateRating and individual Review entries, only where real reviews exist on the page.
• BreadcrumbList so the category hierarchy is explicit.
• FAQPage on sizing, compatibility and policy questions, using the actual questions customers ask.
• Organization on the site root with your legal name, logo, contact points and social profiles.
• ItemList on category and buying-guide pages to enumerate the products in order.

A minimal product example:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Merino Base Layer, Long Sleeve",
  "sku": "ML-2201-NVY",
  "brand": { "@type": "Brand", "name": "Northmoor" },
  "material": "100% merino wool, 190gsm",
  "offers": {
    "@type": "Offer",
    "price": "89.00",
    "priceCurrency": "GBP",
    "availability": "https://schema.org/InStock",
    "itemCondition": "https://schema.org/NewCondition",
    "shippingDetails": { "@type": "OfferShippingDetails" }
  }
}
</script>
      

Keep it truthful. The single most common ecommerce schema failure is markup that drifts out of sync with reality: a cached price that no longer matches the cart, or InStock on a product that sold out last week. Generate structured data from the same source of truth as the page itself, expire it with the page, and never hand-maintain it in a template. Wrong markup is worse than none, because it teaches assistants to distrust your domain.

Feeds and Marketplaces

Merchant feeds and marketplace listings still matter, because they are widely syndicated and easy for machines to consume. Keep titles, attributes, identifiers and imagery consistent between your feed and your own product pages, so that an assistant reconciling multiple sources sees one coherent product rather than three conflicting ones. But do not let the marketplace listing become the richest description of your product. If Amazon or a retail partner carries more detail than your own page, that is the page that gets cited, and the click goes somewhere other than your checkout. Your site should always be the most complete source about your own catalogue.

It is also worth publishing a machine-readable summary of your site for assistants. An llms.txt file points crawlers at your most useful pages, and a broader agent experience approach makes sure automated visitors are served the same substance as humans.

Quick-Start Checklist

1. Audit your top 50 revenue-generating product pages for extractable body text. Count the words that are not navigation, price or button labels.
2. Add a real HTML specification table to every one of those pages, covering dimensions, materials, weight and compatibility.
3. Write a sizing or fit guide for each category where returns are driven by fit, and link it from every relevant product page.
4. Publish a buying guide per major category that names decision criteria and recommends specific products for specific constraints.
5. Create honest comparison pages for your top products, including at least one credible competitor per page.
6. Rewrite shipping, returns and warranty pages into specific, quotable statements with concrete timeframes.
7. Move review text and any accordion content into server-rendered HTML.
8. Implement Product with Offer, AggregateRating, BreadcrumbList, FAQPage, Organization and ItemList, generated from live inventory data.
9. Set up a check that catches price and availability drift between your pages and your structured data.
10. Ask the assistants your own category questions each month and record which sources they name, then track citations so you can tell whether the work is landing.

Start with the categories that carry your margin. A store that fixes twenty product pages properly will see more citation movement than one that adds thin schema across ten thousand. If you want a read on where your catalogue stands today, a free AEO audit is the fastest way to find the gaps.