Why it matters
- A canonical says "this page is a copy, index that one instead". If search engines follow it, this page leaves the results, and the product or article on it with it.
- Google asks that a canonical's target be a duplicate or a near duplicate: its own check is that most of the main text of the copy also appears in the canonical page.
- When the two pages are different, Google may ignore the canonical, and then chooses which page counts by itself.
- It often comes from a template: a product page copied from another with its
<link rel="canonical">, or a plugin that fills the canonical with the wrong product's address.
How to fix
Give the page its own address as its canonical, or remove the tag:
html
<link rel="canonical" href="https://www.example.com/products/khalas-dates/">- Check the template or plugin that writes the canonical: each product, article and category names itself unless it is a copy of another page.
- For a product's variants (sizes, colours), name the main product page, and mark the variants as a
ProductGroupin the structured data.
How we detect
- We ask once for the page the canonical names, when it is not the page itself.
- We judge only where both pages say what they are: each marks one product (
Product) or one article (Article,BlogPosting,NewsArticle…) in its JSON-LD with a name. - The rule fails when the two names differ (neither one's words are all among the other's), and less than half of this page's main text appears in the other page's: their runs of three words, compared after folding Arabic letter forms, case and digits. The main text is the text inside
<main>, or, without it, the text outside the header, the navigation, the footer and the asides. - Not judged: product variants marked as such (
ProductGroup,isVariantOf,inProductGroupWithID,hasVariant), the second and later pages of a list (pagination-canonical), and a canonical that redirects, answers an error or is not indexable (canonical-target-invalid).