Showing 107,000 pages about Core Web Vitals and AI search

As AI-led search becomes a real driver of discovery, old assumptions are back with new urgency. If AI systems reduce the quality from user experience, and Core Web Vitals (CWV) is Google’s most visible proxy for experience, then strong CWV performance should correlate with strong AI visibility.

The logic makes sense.

Faster page load times lead to smoother page load times, increased user engagement, improved signals, and AI systems that reward the result (supposedly)

But the mind is not proof.

To test this properly, I analyzed 107,352 web pages that appeared prominently in the Google AI Overview and AI Mode, examining the distribution of Core Web Vitals at the page level and comparing them to performance patterns in AI-driven search and response systems.

The goal was not to confirm that performance “matters”, but to understand how it matters, where it matters, and whether it meaningfully differentiates itself in the context of AI.

What emerged was not a simple yes or no, but a nuanced conclusion that overshadows existing assumptions about how many groups are prioritizing technology development in the AI ​​era.

Why distribution is more important than scores

Most of Core Web Vitals’ reporting is built around thresholds and averages. Pages pass or fail. Sites are summarized by average scores. Dashboards reduce thousands of URLs to a single number.

The first step in this analysis was to completely depart from that framework.

When Paint’s Largest Content was seen as a spread, a pattern quickly became clear. The dataset showed a heavy right skew.

The median LCP values ​​are clustered in a broadly acceptable range, while the long tail of extreme outliers extends well beyond it. A small percentage of pages were horribly slow, but had a negative impact on average.

Cumulative Layout Shift showed the same problem. The majority of pages recorded a CLS of almost zero, while a few showed significant instability.

Also, the description raised a universal problem that did not reflect the living reality of many pages.

This is important because AI systems don’t think beyond averages, if they think about user engagement metrics at all.

They examine individual documents, templates, and content roles. The broad CWV score is an abstraction created for ease of reporting, not the signal used by the AI ​​model.

Before discussing the relationship, one thing is clear. Core Web Vitals is not a single signal, it is a distribution of behavior among a mixed population of pages.

Communication

Because the data were skewed and not normally distributed, a standard Pearson correlation was not appropriate. Instead, I used Spearman’s rank correlation, which tests whether pages that rank higher in one ranking also tend to rank higher or lower in another, without assuming a linear relationship.

This is important because, if Core Web Vitals were closely linked to AI performance, pages that perform better in CWV would also tend to perform better in AI visibility, even if the link was weak.

I’ve had a few bad relationships. It was there, but it was limited. For Large Content Paint, the correlation ranged from -0.12 to 0.18, depending on how the AI’s visibility was measured. For Cumulative Layout Shift, it was weak as well, usually between -0.05 and -0.09.

This relationship is evident when looking at a large amount of data, but it is not materially strong. Most importantly, they don’t suggest that fast or stable pages are consistently visible to AI programs. Instead, they point to a subtle pattern.

The absence of upside, and the presence of downside

The data does not support the claim that improving Core Web Vitals beyond the basic limits improves AI performance. Pages with good CWV scores did not reliably outperform their peers in AI placement, citations, or retrieval.

However, negative correlations are instructive.

Pages sitting in the extreme tail of CWV performance, especially LCP, were least likely to perform well in AI scenarios.

These pages tend to show low engagement, high bounce, and weak behavioral reinforcement signals. Those second-order effects are exactly the kinds of AI signals that rely on, directly or indirectly, when learning what to trust.

This reveals the true nature of the relationship.

Core Web Vitals does not act as a growth lever for AI visibility. They act as a barrier.

Efficiency does not create profit. Big failure creates inefficiency.

These differences are easy to miss if you only examine pass rates or averages. It becomes clear when examining the distribution and rank-based relationships.

Why ‘passing CWV’ is not a difference

One reason that the positive relationship that many expect does not come easily. Web Critical Passing is no longer a rarity.

In this dataset, most of the pages already meet the recommended limits, especially for CLS. If most people remove the bar, removing it does not separate you. It just keeps you in contention.

AI systems do not discriminate between pages because one loads in 1.8 seconds and the other in 2.3 seconds. They choose between pages because one explains the concept clearly, aligns with established sources, and satisfies the user’s intent, while the other does not.

Core Web Vitals ensures that the experience does not compromise those qualities. They don’t hold.

Redefining the role of Core Web Vitals in AI strategy

That’s not to say that Core Web Vitals aren’t important. It is that their role is misunderstood.

In an AI-led search environment, Core Web Vitals serves as a risk management tool, not a competitive strategy. They prevent pages from entering contention due to negative experience signals.

This reorganization has practical implications for improving AI’s visibility strategy.

Chasing incremental CWV gains on pages that are already accepted is unlikely to bring returns with AI visibility. It leverages the engineering effort without replacing existing AI systems of choice.

Targeting the extreme tail, however, does matter. Pages with really poor performance generate negative behavioral signals that can suppress trust, reduce reuse, and weaken low read signals.

The goal is not to make everything perfect. It’s to ensure that the content you want the AI ​​systems rely on isn’t disrupted by avoidable technical failures.

Why is this important?

As AI systems increasingly mediate acquisition, brands are looking for controllable levers. Core Web Vitals feel attractive because they are measurable, common, and practical.

The risk is to miscalculate impact.

This analysis suggests a more ethical approach. Treat Core Web Vitals like table stakes. Eliminate major failures.

Protect your most valuable content from technical debt. Then return the focus to the things that AI systems use to demonstrate value, such as clarity, consistency, goal alignment, and behavioral validation.

Core Web Vitals: Gatekeeper, not arbiter

Based on an analysis of 107,352 AI virtual web pages, the relationship between Core Web Vitals and AI performance is real, but limited.

There is no strong positive correlation. Increasing the CWV beyond the basic limits does not reliably improve AI visibility.

However, a measurable negative relationship exists in extremes. Major performance failures are associated with poor AI results, mediated by user behavior and engagement.

Therefore, Core Web Vitals is best understood as a gateway, not a beauty signal.

In an AI-led search environment, this clarity is essential.