A shopper asks ChatGPT for a waterproof jacket under $200. You sell one with 400 reviews, but they are spread across 48 colour and size pages. An engine that reads one of those pages sees about eight. A competitor with one page and its full review count looks like the stronger pick. Variant sprawl is one product published as dozens of near-identical pages, with its reviews, links and facts split between them. We fix it by moving every variant under one canonical parent page, with each variant's details in structured attributes.
What Is Variant Sprawl, and Why Does It Matter for AI Recommendations?
Sprawl happens when a product with six colours and eight sizes is published as 48 separate product pages. Each page carries the same description and its own slice of the reviews. An engine that retrieves passages sees 48 near-copies and no clear owner of the product.
According to NielsenIQ, nearly 75% of shoppers use AI during product discovery. An AI answer names a few products and gives a reason for each. That reason is evidence: a rating, a review count, or an attribute that fits the request. The engine can only quote what is on the page or feed record it retrieves. With variant sprawl, that record holds a fraction of your reviews, so your product competes on a fraction of its evidence.
Google states the general principle in its canonicalization guide. Choosing a canonical URL consolidates "the signals they have for the individual URLs (such as links to them) into a single, preferred URL". That applies directly to Google's AI Overviews and AI Mode. A page must be indexed and eligible to be shown in Google Search with a snippet to appear there as a supporting link. We have not found an equivalent statement from ChatGPT, Perplexity or Claude, so for those we apply the same logic. Evidence on one page can be found and quoted, and evidence divided across 48 pages is harder to find.
How Does Sprawl Split the Signal?
Sprawl divides three things: reviews, links and the product's identity.
Reviews. In the jacket example, each of the 48 pages shows about eight of the 400 reviews. A shopper, or an engine, reading one page sees a thin record for a product that has a strong one. Google's structured data guide supports a group-level `aggregateRating` that is "representative of all variants".
Links. Internal and external links point at whichever variant URL a merchandiser, an affiliate or a blogger happened to use. Consolidation sends them all to one URL.
Identity. When 48 pages share one description, an engine has to decide which is the product. Google tells merchants to use the same item group ID for all variants of the same product. The value should be the parent SKU, which marks the product as one item with options.
Regional duplicates split the same signals across markets, but they need a different fix. Keep each market's page, since each carries its own price and availability. Link them with hreflang annotations instead of pointing them at one global canonical.
How Do You Consolidate Variants?
The fix has four parts, in this order:
- Pick one canonical parent page per product. Google's ecommerce guidance covers variants identified by optional query parameters. It says to use the URL with the query parameter omitted as the canonical URL. It also says each variant should be identifiable by a separate URL, which a parameter provides.
- Put variant details in structured attributes. Use `ProductGroup` with `variesBy` and `hasVariant` in the markup, so Google can tell which products are variations of the same parent. Colour, size and material become fields Google can read as variant attributes.
- Group the feed. Use `item_group_id` on every variant. In ChatGPT's feed, a shared `group_id` does the same job, and OpenAI says it is what establishes a variant group. Google's `item_group_title` attribute says to keep the group title shorter and more generic than the titles of variant products, without size or colour.
- Move the reviews and copy to the parent. Show the full review count and description on the canonical page. Render the reviews in raw HTML, as covered in our article on review widgets.
Google's guide lists `rel="canonical"` as a strong signal that the specified URL should become canonical. Google can still choose another URL if your internal links, redirects and sitemap disagree, so check that they all name the same one.
Should Every Variant Be Merged?
No. Keep a separate variant page only when shoppers search for that variant by name. The page also needs content of its own, such as a named colourway. Sizes rarely qualify, because a size is an attribute of the product.
Feed titles work differently from page titles. OpenAI's product feed spec asks for a product name "including the selected variant when relevant". The variant detail therefore belongs in each item's feed title. Only Google's feed uses a separate group title; OpenAI lists `item_group_title` as legacy and not used for discovery.
Who Owns the Fix?
If you run an enterprise catalog, split the work by owner before you start. Canonical logic, URL patterns and the review widget usually sit in a global template, so they need an HQ ticket. Feed grouping and parent titles can usually move locally.
If you are a small team, you can do all four steps in a week for your top products. Start with the products that have the most reviews, because those lose the most when split.
FAQs
What is variant sprawl?
Variant sprawl means one product is spread across many near-duplicate pages, usually one per colour or size. The pages share a description and divide the reviews, links and product facts between them.
Does a canonical tag guarantee that AI cites the parent page?
No. A canonical tag tells Google which URL you prefer, and Google can override it when other signals disagree. Outside Google, we have not found AI engine documentation on canonical tags. We treat consolidation as the way to make the evidence easy to find.
Should I redirect old variant URLs?
Where an old standalone variant page has no demand of its own, redirect it permanently to the parent with that variant preselected, such as `/jacket?colour=red`. Google says a 301 redirect is used as a signal that the redirect target should be canonical. Its variant markup needs each variant reachable at a distinct URL. A temporary redirect does not carry the canonical signal.
How do I check whether I have variant sprawl?
Export your product list and count how many separate product records share one base name. If a product has more than one record per colour or size, and each shows only part of the reviews, you have it.
The Takeaway
Consolidate the URLs, mark up the group, and give every variant the same group ID in your feeds.
Consolidation is one part of making a catalog visible to AI engines. Once each product has one complete page, that page needs to match the situations shoppers describe to AI. We cover that step in our guide to use-case clustering.
Our GEO for ecommerce service does this work for you. A specialist on our team consolidates your variants, clusters your catalog by shopper intent, and tunes your feed attributes and product schema. We then track whether AI engines start recommending your products. To see where they stand today, get a free AI visibility assessment.