Core Web Vitals: Which version truly impacts your ranking when Google shows the AMP?

Why does this precision about the displayed version change the game?

Most SEOs instinctively think that ranking signals come from the canonical URL or the site's 'main' version. Mueller's statement shatters this misconception. When Google decides to display an AMP version in the SERPs, it is that version that provides the CWV data for ranking. Not the classic desktop version that you spent three months optimizing.

This principle extends to international sites with distinct URLs. If you have example.com/fr and example.com/en, and Google displays the French version, it is the metrics of that French URL that count. Not those of a hypothetical ‘reference’ version. This logic seems coherent, but it seriously complicates technical management.

What does this mean for current AMP strategies?

Many sites have abandoned AMP or are poorly maintaining it, focusing on the classic mobile version. If Google continues to serve AMP in certain contexts (news, carousels), you are left with a degraded version that negatively impacts your ranking. The CWV metrics of this neglected AMP page become your Achilles' heel.

The opposite also exists: some sites have an ultra-optimized AMP but a mediocre main version. If Google switches to the non-AMP version, ranking suffers. Consistency between versions becomes a strategic issue, not just a matter of technical comfort.

How does Google determine which version to display?

Mueller's statement does not delve into this detail, and that’s where the issue lies. Google has criteria for choosing AMP vs non-AMP: perceived speed, search context, device type, eligibility for specific features (Top Stories, carousels). But these criteria are neither fixed nor exhaustively documented.

For international sites, it’s even more opaque. Hreflang plays a role, user location does too, but Google may decide to serve a different URL than expected. You only have partial control over this decision, making optimization frustrating.

• The displayed version in the SERPs determines the CWV metrics used for ranking, not the canonical URL.
• AMP sites must optimize this version just as much as the main one, or risk penalizing their ranking.
• For international sites, each country variant URL must meet CWV standards, even if you think another version is ‘main’.
• Google retains control over the choice of the displayed version, with not entirely transparent criteria.
• Neglecting a version on the pretext that it generates little direct traffic can become a strategic mistake if Google serves it in the results.

Is this statement consistent with what we observe on the ground?

Yes and no. On news sites where AMP is actively used, we do see performance disparities between the two versions reflected in the ranking. When AMP is poorly optimized, positioning on certain queries drops, even if the desktop version is impeccable. This aligns with Mueller’s logic.

However, for international sites with distinct URLs, it’s less clear. Many SEOs report inconsistencies: Google sometimes seems to use mixed signals between versions, especially when hreflang is misconfigured. [To verify]: does Google partially aggregate CWV data from multiple variants, or does it strictly applies the rule of 'displayed version only'? The data is lacking to make a determination.

What are the gray areas in this statement?

Mueller does not specify what happens when Google frequently switches between versions for the same user or query. If half the time it’s the AMP, and the other half the classic mobile version, which version prevails for ranking? One might assume it is dynamic, but there’s no confirmation.

Another unclear point: the CRuX data (Chrome User Experience Report), which feeds into CWV, is aggregated over rolling 28 days. If you fixed your AMP a week ago, the old metrics still weigh in. How long does it take for the improvement to reflect in the ranking? Mueller does not specify. In practice, expect a minimum of 4 to 6 weeks.

In what cases might this rule not apply strictly?

Let’s be honest: Google has always used hybrid signals when the data is incomplete. If AMP lacks sufficient Chrome traffic to generate reliable CRuX metrics, Google might revert to the main version. But this is a hypothesis — nothing official.

For very large sites with dozens of international variants, we sometimes observe anomalous behavior: a minor version (for example, the French-speaking .be) inheriting signals from a larger version (.fr). Is this a bug, a voluntary aggregation, or a side effect of URL clustering? Impossible to say without feedback from Google.

What should be prioritized in an audit of your site?

The first step: identify all the URL versions that Google potentially indexes and displays. AMP, international versions, mobile variants with distinct domains. Use Search Console to find out which ones actually appear in the SERPs. Do not rely on your intuition — Google sometimes serves versions you thought were secondary.

Next, measure the CWV of each version individually. PageSpeed Insights, CRuX API, or the 'Core Web Vitals' tab in Search Console. Compare with your standards. If the AMP shows an LCP of 4 seconds while your classic mobile stands at 1.5 seconds, you have a problem — it’s potentially that poor AMP version that’s weighing down your ranking.

How to prioritize fixes when managing multiple versions?

First, focus on the versions that Google actually displays for your strategic queries. An AMP that never appears in the SERPs can wait. On the other hand, an international version capturing 20% of your impressions on high ROI keywords deserves immediate optimization.

For international sites with distinct URLs, create a prioritization matrix: impression volume × CWV gap × business value. This gives you a rational attack order. Do not fall into perfectionism — optimizing 15 mediocre versions takes time. Better to start with the 3 that yield results.

What mistakes should you absolutely avoid in this context?

Classic mistake: disabling AMP without verifying that Google has indeed switched to the classic mobile version in the results. If Google continues looking for AMP and encounters a 404, your ranking collapses. Make the transition cleanly: first ensure that the non-AMP mobile version appears, and only after, remove the AMP.

Another trap: believing that hreflang is enough for Google to serve the correct international version. Hreflang is a signal, not a directive. If your .de version has catastrophic CWV, Google may prefer to display the .com version with a language warning. Result: it’s the .com version that counts for ranking on German queries, and you lose control.

• List all indexed URL versions (AMP, international, mobile variants) and check which ones actually appear in the SERPs
• Measure the Core Web Vitals of each version individually via Search Console or CRuX API
• Identify the versions displaying a significant CWV gap with your standards (LCP > 2.5s, CLS > 0.1, INP > 200ms)
• Prioritize optimizations based on impression volume × CWV impact × business value of each version
• Never disable an AMP or international version without validating that Google indeed displays the desired alternative in the results
• Automate CWV monitoring by version to quickly detect regressions (weekly alerts minimum)