GEO for ecommerce: getting product pages cited by ChatGPT and Gemini

Shopping queries are one of the fastest-growing slices of AI search. People ask ChatGPT for 'the best running shoes for flat feet under £120', they ask Gemini to compare two espresso machines, and they ask Perplexity for vegan protein powders with no artificial sweeteners. The engine reads product pages, category pages and reviews on the open web and synthesises a recommendation – with citations. If your store is not in those citations, you are invisible in the channel that is quietly cannibalising classical product search.

Why ecommerce is harder than content GEO

Most ecommerce templates were built for human shoppers and Googlebot, in that order. They lean on client-side rendering, image-only specs, lazy-loaded reviews and JavaScript-driven variant pickers. AI crawlers see almost none of that. The fix is not to rebuild your store – it is to make sure every claim a shopper might care about is present, in text, in the initial HTML.

The five things every product page needs

1. A server-rendered Product JSON-LD block with name, brand, sku, gtin, price, priceCurrency, availability, aggregateRating and review.
2. A plain-text spec list – not an image, not a PDF. Material, dimensions, weight, compatibility, ingredients, certifications.
3. A short, factual description that answers the obvious shopper question in the first 200 characters.
4. Real customer reviews rendered in the HTML, not injected by a third-party widget after load.
5. Clear shipping, returns and warranty text on the page itself – not hidden behind a tab that loads on click.

Category pages are citation goldmines

AI engines love comparative content. A well-structured category page – with a short intro, a comparison table, and clear criteria for why each product is on the list – is far more likely to be cited than an individual product page. Treat your top categories like editorial buyer's guides.

• Add a 150–250 word intro explaining who the category is for and what to look for.
• Include a comparison table with the same attributes across every product.
• Use real H2 headings for each product, not styled divs.
• Link to the deeper product page for full specs.

Schema that actually moves the needle

Three schemas do most of the work for ecommerce GEO: Product, Offer and Review (with AggregateRating). Add BreadcrumbList for navigation context, and Organization on your homepage so the engine knows who is selling. Validate every page with Google's Rich Results test – if it fails there, AI parsers will also struggle.

{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Acme Trail Runner v3",
  "brand": { "@type": "Brand", "name": "Acme" },
  "sku": "ACM-TR3-BLK-10",
  "description": "Lightweight trail running shoe with carbon plate and Vibram outsole.",
  "offers": {
    "@type": "Offer",
    "price": "139.00",
    "priceCurrency": "GBP",
    "availability": "https://schema.org/InStock"
  },
  "aggregateRating": {
    "@type": "AggregateRating",
    "ratingValue": "4.6",
    "reviewCount": "218"
  }
}

Reviews: the single biggest unlock

Review content is what AI engines quote when they recommend a product. If your reviews live inside a JavaScript widget from a third-party platform, the crawler sees an empty div. Ask your reviews provider for a server-rendered fallback, or proxy the review HTML through your own server. The lift is usually one engineering ticket and it is the single biggest GEO unlock for most stores.

What to stop doing

• Hiding specs inside tabs that render on click.
• Putting key claims in hero images with no alt text.
• Blocking GPTBot, ClaudeBot or PerplexityBot in robots.txt by default – you are opting out of the channel.
• Auto-generating thin variant pages with duplicated content.

A 2-week plan

1. Week 1: audit your top 20 products, add Product + Offer + AggregateRating JSON-LD, move specs and reviews into server-rendered HTML.
2. Week 2: rewrite your top 5 category pages as buyer's guides with comparison tables, then check AI bot access in robots.txt.

Run a free audit on a key product page first – it will tell you within 60 seconds whether AI engines can actually read what you sell.