Reading the Search Console statuses
The Page indexing report files its canonical states under reasons pages are not indexed, and that framing leads people to treat them as failures. Most are Google working as intended. Two facts sort every status: whether you declared a canonical at all, and whether Google accepted it. URL Inspection shows both side by side, as the user-declared canonical and the Google-selected canonical, and that comparison is the first thing to check before changing any markup.
- Informational. Alternate page with proper canonical tag means the page points elsewhere and Google agreed. Large counts are normal on sites with parameter, print or mobile versions. It is a problem only when a page you want indexed lands here because a template pointed it away from itself.
- Worth reviewing. Duplicate without user-selected canonical means Google found duplicates, you stated no preference, and Google chose for you. If its choice is the URL you would have picked, nothing is lost; if not, declare one.
- Investigate. Duplicate, Google chose different canonical than user means you declared a canonical and Google overrode it, usually because the target is weak or another signal disagrees.
- Investigate. Duplicate, submitted URL not selected as canonical is best read as the sitemap-submitted counterpart of the previous status. The usual trigger is a sitemap listing URLs that are not canonical.
An override is a prompt to look, not a verdict against the site. Google is often right: it may have preferred the HTTPS version, the one with more internal links, or the one your own redirects already favor. Decide first whether you actually disagree with its choice. If you do, the fix is to put every signal behind the preferred URL, so that the tag, the sitemap, the internal links and the redirects all name the same address, rather than restating the tag and waiting.
How conflicts arise
Conflicts come from signals that contradict each other, usually because two systems each believe they own the canonical. The most mechanical case is multiple canonical tags on one page, typically a theme and an SEO plugin each printing its own, which may lead Google to ignore every one of them and fall back on its own choice. A subtler case begins when a developer adds a robots rule to a page whose canonical already points elsewhere. Canonical and noindex together ask Google to consolidate the page and drop it at the same time, and the noindex rule tends to win.
Other conflicts are logical rather than mechanical. Canonical tags and pagination go wrong when a template points pages two, three and four at page one, which declares the deeper pages duplicates even though each lists different items and links to different products. A related failure is the non-indexable canonical: a target that redirects, returns a 404, carries a noindex rule or is blocked in robots.txt asks Google to consolidate onto a URL it cannot index. Point every canonical straight at a final URL that returns 200.
A cross-domain canonical is technically supported as a hint, but for syndicated articles Google now prefers that partners keep their copy out of the index rather than point it back at the original, because republished pages often differ too much for the hint to hold. The self-referencing canonical sits at the other end of the scale: the baseline that prevents much of this, because every variant of a page arrives with an unambiguous answer attached. Its own failure is subtle, a template that self-references the variant being served, parameters and all, instead of the clean URL.
Where duplicate URLs come from
The third group is a supply problem: the URLs that make canonicalization necessary. Host and protocol variants are the oldest source, and settling them starts with a canonical domain: one host, with or without www, served over HTTPS. Search Console retired its preferred domain setting in 2019, so the choice now rests on redirects, internal links and consistent canonicals that all name the same host and protocol. Canonical redirects enforce the choice with a single 301 or 308 hop from every variant. Chains, loops and platforms that default to a temporary 302 are the defects that turn up most.
Parameters and variants multiply URLs faster than hosts do. URL parameters for tracking, sorting and faceted navigation create near-identical pages, and a canonical consolidates them for indexing without stopping Google from crawling them, which is why crawl control is a separate decision with a separate set of tools. Product variants have two valid models: one URL per color or size, each canonicalizing to itself, or one base URL that every variant consolidates to. Either works when it is applied consistently across the catalog.
Some duplicates are paired by design. AMP pages carry a canonical pointing at the regular page, and m-dot mobile URLs point theirs at the desktop URL, with a matching annotation in the other direction; if either half is missing, Google may ignore the pair. Files with no HTML head, such as PDFs, can declare a canonical in an HTTP Link header instead. That header is invisible in page source, so it is easy to miss when an HTML tag on the same resource says something different.
Finally, non-canonical URLs are normal, and every site has them. Audit tools flag them because they turn up where they do not belong: in XML sitemaps, in internal links, in hreflang annotations, and as the targets of other canonicals. A duplicate address existing is not the defect. Pointing the site's own signals at it is, because each of those references is a vote for a URL you have asked Google not to choose.
A sensible order of work
When an audit returns a long list, fix whatever makes the canonical unreadable first: several tags on one page, tags outside the head, and targets that redirect or fail. Next, remove contradictions, such as noindex paired with a canonical or paginated pages pointed at page one. Then tighten supply with redirects and parameter handling. Only after that is it worth reading the remaining Search Console statuses, because by then most of them describe choices you made deliberately.