Google picked a different canonical than yours: check the signals in order

The signals, and how much weight each has

Google also says none of these methods is required. A canonical is therefore a request that Google weighs against everything else it sees. Search Console records the outcome in the Page indexing report: "Duplicate, Google chose different canonical than user" means the page is marked as canonical for a set of pages but Google thinks another URL is better; "Page with redirect" is a non-canonical URL that redirects to another page. The URL Inspection tool shows the user-declared canonical and the Google-selected canonical side by side.

Six checks, in the order that finds the cause

1. Does the canonical URL return 200 itself? Request it without following redirects. If it answers 301, the tag points at a redirect.
2. Does the page's served URL equal its canonical? After redirects, the URL you land on should be the one the tag names.
3. Does the sitemap list the canonical form? A sitemap full of URLs that redirect is a list of non-canonical URLs.
4. Do internal links use the canonical form? Linking to a variant votes for the variant.
5. Are there near-duplicates? Two pages with the same content under different URLs are the case canonicals exist for; see the Crawled guide for merging them.
6. Which variants exist? Check HTTP and HTTPS, www and non-www, trailing slashes, /index.html, upper and lower case, and tracking parameters, and make each redirect to one form.

A real case: this site

While writing this guide the checker below was run against this site's live sitemap on 30 October 2026. The host (Cloudflare) answers /seo/seo-audit-checklist.html with a 301 to /seo/seo-audit-checklist, and serves the clean URL with a 200. Every page, however, declared the .html URL as its canonical, listed the .html URL in the sitemap, and linked to it internally.

That is the contradiction in its purest form: each page's canonical named a URL that redirected back to the page that declared it. Nothing in Google's documentation says how it resolves a loop like that, and I have no Search Console data showing which URL it picked, so the effect is untested. The signals were inconsistent, and that is reason enough to fix them.

The fix applied was to make the canonical, og:url, structured-data URLs, sitemap and llms.txt all use the clean URL the host serves. The checker was then run against the fixed files, served with the same redirect rules on a local test server: 0 of 51 sitemap URLs had an issue. The live site has not yet been redeployed with that change as this was written, so the "before" numbers above describe what is live.

The checker script

It reads a sitemap and checks each URL. Run it with the command below:

python3 canonical_check.py https://example.com/sitemap.xml

import re, sys, time, urllib.request, urllib.parse, urllib.error

class NoRedirect(urllib.request.HTTPRedirectHandler):
    def redirect_request(self, *args, **kwargs):
        return None

raw = urllib.request.build_opener(NoRedirect)
UA = {'User-Agent': 'canonical-check/1.0'}

def status(url):
    # status and Location of one request, redirects not followed
    try:
        with raw.open(urllib.request.Request(url, headers=UA), timeout=20) as r:
            return r.status, None
    except urllib.error.HTTPError as e:
        return e.code, e.headers.get('Location')

def page(url):
    # final URL and HTML after following redirects
    with urllib.request.urlopen(urllib.request.Request(url, headers=UA), timeout=20) as r:
        return r.geturl(), r.read().decode('utf-8', 'replace')

sitemap = urllib.request.urlopen(urllib.request.Request(sys.argv[1], headers=UA), timeout=20).read().decode()
urls = re.findall(r'<loc>([^<]+)</loc>', sitemap)
problems = 0
for u in urls:
    issues = []
    code, loc = status(u)
    if code in (301, 302, 307, 308):
        issues.append('sitemap URL redirects (%s) to %s' % (code, urllib.parse.urljoin(u, loc or '')))
    try:
        final, html = page(u)
    except Exception as e:
        print('ERROR  %s  %s' % (u, e)); problems += 1; continue
    m = re.search(r'<link[^>]+rel=["\']canonical["\'][^>]*href=["\']([^"\']+)', html)
    canon = urllib.parse.urljoin(final, m.group(1)) if m else None
    if not canon:
        issues.append('no rel=canonical')
    else:
        if canon.rstrip('/') != final.rstrip('/'):
            issues.append('canonical %s differs from the URL served (%s)' % (canon, final))
        ccode, cloc = status(canon)
        if ccode in (301, 302, 307, 308):
            issues.append('canonical URL itself redirects (%s) to %s' % (ccode, urllib.parse.urljoin(canon, cloc or '')))
        elif ccode != 200:
            issues.append('canonical URL returns %s' % ccode)
        if canon not in urls:
            issues.append('canonical is not listed in the sitemap')
    if issues:
        problems += 1
        print('FIX    ' + u)
        for i in issues: print('         - ' + i)
    time.sleep(0.1)
print('\n%d of %d sitemap URLs have at least one issue' % (problems, len(urls)))

On a local test server with a planted bad canonical and a redirecting sitemap URL, it flagged exactly those two. It only reads pages and sends a request or two per sitemap URL, with a short pause, so it is light enough to run on a small site.

What to fix

If URLs have changed as part of a move, the redirect-map guide has a script for the redirect side, and the SEO audit checklist lists canonical conflicts among its crawlability checks.

What this does not cover