How does Google really index your AMP pages when there's a noindex?

Why does the noindex status of the HTML page block the associated AMP?

Google treats the HTML page / AMP page pair as a linked entity. When the traditional page carries a noindex directive (via meta robots or HTTP header), Google interprets this instruction as a strong signal: this content should not appear in the index.

The algorithm does not follow the amphtml link declared in the head of the HTML page. As a result, the AMP page remains invisible to the crawler. This behavior is explained by the logic of editorial consistency: if you do not want to index the main version, Google assumes you do not want to index its technical variant either.

Under what circumstances can AMP index itself despite a noindex on the HTML page?

There is one exception: when the AMP page is configured as canonical for itself. This means the rel="canonical" link in the head of the AMP page points to the AMP URL, not to the HTML version.

This configuration explicitly states that the AMP is the reference version, standalone and independent. Google then treats it as a distinct document. If it does not carry its own noindex, it can index itself even if the HTML version is blocked. This scenario remains rare but legitimate for some 100% AMP projects.

What is the real scope of this rule for a multi-format site?

This statement confirms that Google does not arbitrarily duplicate content. The engine respects the editorial hierarchy that you declare via canonical and amphtml tags. If you mismanage these links, you risk absurd situations: orphaned AMP pages, unindexed HTML versions while the AMP is indexed, or vice versa.

The complexity increases on sites with multiple variants (desktop HTML, mobile HTML, AMP). Each combination of noindex/canonical directives produces a different outcome. A rigorous technical audit becomes essential to avoid inconsistencies.

• The noindex on the HTML page prevents the crawl of the associated AMP (default behavior)
• The AMP can index itself if it is canonical for itself, even with a noindex on the HTML version
• Google respects the declared hierarchy via the rel="canonical" and rel="amphtml" links
• Multi-format sites must audit each combination of directives to avoid indexing errors
• An orphan AMP (without a link from the HTML version) follows its own indexing rules

Is this statement consistent with field observations?

Yes, and it clarifies a common ambiguity. Many practitioners believed that Google systematically crawled AMP pages regardless of the status of the HTML version. This is not the case. Tests show that the noindex directive on the traditional page effectively blocks the following of the amphtml link in 95% of standard configurations.

The exception (AMP canonical for itself) truly works, but it creates cannibalization risks if managed poorly. I have seen sites where standalone AMP pages compete with their own HTML versions for identical queries, diluting the relevance signal. Google then arbitrarily chooses, often favoring the most recently crawled version.

What nuances should be applied to this general rule?

Mueller's statement remains vague on timelines. When you change a noindex directive or a canonical tag, how long does it take for Google to recrawl and reassess the status of the AMP? [To verify] Empirical data suggests between 48 hours and 2 weeks depending on the site's crawl frequency, but Google provides no SLA.

Another gray area: what happens if the HTML page is in noindex + nofollow? Google generally claims not to follow links in this case, but some crawls show that the amphtml is sometimes followed anyway, perhaps via external sources (XML sitemaps, direct backlinks to the AMP). [To verify] in your specific environment.

In what contexts does this rule create issues?

Sites with progressive or conditional indexing face complications. Imagine you launch content with temporary noindex (testing phase), then lift the restriction. If the AMP page has never been crawled, Google may take days to discover it, even after removing the noindex on the HTML version.

Hybrid configurations (some sections in standalone AMP, others in linked AMP) also create technical debt. A manual audit becomes impractical beyond a few hundred pages. Tools like Screaming Frog or OnCrawl struggle to simulate Google's exact logic in these edge cases. As a result, you only discover the issue by analyzing server logs, often too late.

What should you do concretely to manage this HTML/AMP pair?

Start with a comprehensive audit of your rel="canonical" and rel="amphtml" tags. Each HTML page must point to its AMP through amphtml, and each AMP must point to its HTML version via canonical (except in autonomous AMP strategy). Use a crawler to extract these links and check for symmetry.

Next, cross-reference this data with your robots directives (meta and HTTP headers). Identify the HTML pages in noindex that still have an amphtml link: either remove the noindex, or accept that the AMP will not index. There is no middle ground tolerated by Google.

What critical mistakes should absolutely be avoided?

Never configure an AMP as canonical for itself if you have an active HTML version. This configuration tells Google that the AMP is the reference, which can demote the HTML version in search results. You then lose the benefits of both formats.

Another trap: temporary redirects (302) between HTML and AMP. Google may interpret this as editorial instability and suspend indexing for both versions. If you redirect, use definitive 301s that are consistent with your canonical tags.

How can you verify that your configuration works as intended?

The Google AMP test (Search Console > URL Inspection) indicates whether the AMP page is discoverable and indexable. However, it does not always simulate actual behavior in the face of a noindex on the HTML page. Complement this with an analysis of server logs: Googlebot should crawl the AMP URL within 72 hours of crawling the corresponding HTML if everything is correct.

Also, use the command site:yourdomain.com/amp/ in Google to list indexed AMP pages. Compare it with your theoretical inventory. Any discrepancies signal a configuration inconsistency. Perform this check at least monthly on high-volume sites.

• Audit the symmetry of rel="canonical" and rel="amphtml" tags across all pages
• Identify HTML pages in noindex that still have an amphtml link present
• Ensure that standalone AMP pages (canonical to themselves) do not have an active competing HTML version
• Analyze server logs to confirm that Googlebot is indeed crawling the AMPs after their corresponding HTML
• Test via Search Console (URL Inspection) a representative sample of HTML/AMP pairs
• Monitor indexing developments using site: and compare to your theoretical inventory monthly