How does Google actually segment your Core Web Vitals by page groups?

Why doesn't Google measure Core Web Vitals globally?

Contrary to what many practitioners believe, Google does not calculate a single Core Web Vitals score for an entire site. The engine uses data from the Chrome User Experience Report and performs automated segmentation into groups of similar pages.

What does this mean in practice? An e-commerce site will have its product pages evaluated separately from its category pages, which will also be distinct from the blog. Each cluster receives its own CWV metrics, independently of the other sections. This approach allows for a more nuanced assessment of the actual user experience.

How does the grouping of pages work?

When Google analyzes a specific URL for ranking, it first identifies the most appropriate group to which that page belongs. The algorithm looks for patterns of similarity: HTML structure, content type, user behavior, technical architecture.

The exact criteria for grouping are not public — surprising, right? But field observations show that common templates, similar URL depths, and navigation patterns play a role. A product page is rarely in the same group as a blog post, even if both are technically on the same domain.

What's special about AMP pages in this segmentation?

John Mueller points out that if AMP pages form a separate group, their Core Web Vitals metrics do not contaminate the other sections of the site. This is an important clarification: your ultra-fast AMP pages will not magically boost the scores of your standard pages.

Conversely, poorly optimized AMP pages will only impact their own cluster. This isolation of performance by group changes the game for optimization strategies — there's no need to uniformly prioritize the entire site if some groups are already performing well.

• Automated Segmentation: Google creates clusters of similar pages without manual intervention
• Isolated Metrics: each group has its own independent CWV scores
• Precise Targeting: a URL is evaluated based on the performance of its specific group
• AMP Independence: AMP pages do not share their metrics with canonical pages
• Selective Optimization: no need to uniformly improve the entire site if certain groups are already performing well

Does this group approach align with field observations?

Automated segmentation indeed explains why some sections of a site can rank well while others stagnate, despite identical technical infrastructures. CWV audits often reveal significant disparities between page types.

What’s the catch? Google provides no visibility on how these groups are formed or their exact scope. The Search Console displays data by origin (complete domain), not by cluster. [To be verified]: official tools do not allow visualization of this internal segmentation, which seriously complicates targeted optimization.

What areas of uncertainty remain in this statement?

Mueller remains intentionally vague about grouping criteria. SEOs observe that URL depth, templates, and traffic volume seem to play a role — but nothing is officially documented. This opacity poses a problem for large sites with complex architectures.

Another critical point: what is the minimum size of a group to be evaluated? If a section receives few Chrome visits, its CrUX data may be absent or insufficient. In these cases, does Google use metrics from the complete domain, the parent group, or does it penalize by default? A mystery. [To be verified]: the exact behavior on low traffic pages remains undocumented.

Do isolated AMP metrics really change the game?

Let's be honest: AMP is largely abandoned since Google removed its competitive advantage in Top Stories. This clarification on the isolation of AMP groups comes late and now concerns only a marginal fraction of sites.

That said, the underlying principle remains valid: if you maintain alternative versions (dedicated mobile, PWA, specific subdomain), their CWV performance does not impact the rest of the site as long as they form a distinct group. This is rather reassuring for progressive migrations or A/B tests on subsets of URLs.

How can you concretely identify the page groups on your site?

Google provides no official tool to visualize this segmentation. The only workable approach is to analyze CrUX data by page type and cross-reference with template patterns. Use BigQuery to access the complete CrUX datasets and manually segment by URL prefixes, depths, or HTML structures.

Cross-reference this data with server logs and Search Console reports to identify clusters that receive sufficient Chrome traffic. Sections with fewer than 1000 monthly visits are likely to lack reliable data — and thus be evaluated differently. Also test via PageSpeed Insights on multiple URLs of the same type to detect scoring disparities.

Should you optimize the entire site uniformly or prioritize certain groups?

Segmentation radically changes the optimization strategy. There's no need to refine the CWV of a blog that is already performing well if your product pages are in the red. First, identify the critical groups for your business — those that generate conversions and revenue.

Prioritize the clusters that combine high traffic and poor performance. A group of 50 pages visited 100,000 times/month deserves more attention than 5,000 pages with a total of 500 visits. However, be cautious: a strategic group with low Chrome traffic may lack CrUX data and switch to default metrics — a scenario to absolutely avoid.

What to do if your AMP pages or alternative versions drag their own group down?

If you're still maintaining AMP (or mobile subdomains, dedicated PWAs), check their specific CWV metrics. Isolation means they do not impact the rest of the site, but they can penalize their own visibility if their scores are poor.

Two options: either you optimize these alternative versions to meet CWV thresholds, or you completely abandon them. Maintaining AMP in 2025 without a clear strategic reason (news, specific partnerships) is generally a waste of time and resources. Instead, consolidate onto a single high-performing version.

• Extract CrUX data via BigQuery and segment by page types or URL prefixes
• Identify high-traffic groups with poor CWV scores (LCP > 2.5s, CLS > 0.1, INP > 200ms)
• Prioritize optimization on strategic clusters (conversion, revenue, key organic traffic)
• Ensure each critical group receives sufficient Chrome traffic to generate reliable CrUX data
• Specifically audit AMP pages or alternative versions if you're still maintaining them
• Test improvements on a representative sample of each group before global deployment