Does PageSpeed Insights truly measure the real performance of your site?

What is the difference between lab data and field data?

Before this integration, PageSpeed Insights primarily relied on synthetic lab tests: a controlled browser, a calibrated network, reproducible conditions. The issue? These tests did not capture the diversity of real experiences. A site might perform excellently in lab conditions on fiber but struggle on 4G from a low-end smartphone.

With the Chrome User Experience Report, Google aggregates metrics from millions of real Chrome users. This data reflects variable network latency, diverse devices, and installed extensions. It shows the speed as experienced by your visitors, not the speed you get on your MacBook Pro connected via Ethernet.

Which CrUX metrics are displayed in PSI?

The tool now shows three key metrics from CrUX: First Contentful Paint, Largest Contentful Paint, and Cumulative Layout Shift. Each is accompanied by a distribution (good / needs improvement / poor) calculated over the previous rolling 28 days. This distribution determines whether your origin passes the Core Web Vitals or not.

If your site doesn't have enough Chrome traffic to appear in CrUX, PSI reverts to synthetic data only. In other words: no real data, no validated Core Web Vitals label. For a low-volume site, this is a blind spot.

Why does this evolution change the game for SEO?

Google has clearly stated that the Core Web Vitals are a ranking signal. These metrics come from CrUX. By measuring real performance rather than theoretical metrics, PSI becomes a reliable proxy for what Google sees from a ranking perspective. Ignoring CrUX data is like driving blind.

This logic mandates a change in approach: optimizing for a lab score is no longer sufficient. You need to understand how your real users load your pages, identify problematic segments (3G mobile, remote areas from your CDN), and fix bottlenecks that penalize the majority.

• CrUX Data: aggregation over a rolling 28 days from millions of real Chrome users
• Exposed Metrics: FCP, LCP, CLS with distribution good/average/poor
• Traffic Threshold: origin absent from CrUX if volume is insufficient → no Core Web Vitals validation
• Ranking Impact: CrUX Core Web Vitals are an official ranking signal
• Lab Limitations: synthetic tests do not reflect variable network/hardware conditions of the field

Is this statement consistent with field observations?

Yes, and it is verifiable. Since the CrUX integration, there is a stronger correlation between PSI scores and rankings on competitive queries. Sites that scored 95/100 in the lab but faltered in real-world conditions saw their Core Web Vitals label turn red. The reality of the field has caught up with cosmetic optimizations.

However, there remain edge cases. Some ultra-targeted B2B sites, with primarily desktop traffic on corporate networks, report flattering CrUX data that does not reflect the general mobile experience. Conversely, a site with high traffic from areas with degraded connectivity may be penalized even if its code is flawless. [To be verified]: Google has never communicated about any potential geographical weighting or network type in calculating the ranking signal.

What nuances need to be added to this announcement?

Google presents this evolution as a step forward in transparency, which is legitimate. However, the CrUX is not free from biases. It only captures connected Chrome users who have accepted data sharing. Safari iOS, Firefox, exotic browsers: all this traffic remains invisible. In certain markets (China, Russia), Chrome is a minority, skewing representativeness.

Another point: the 28-day rolling window smooths variations but introduces inertia. If you deploy a major optimization today, you will have to wait several weeks before CrUX—and hence PSI—fully reflects the improvement. This latency can confuse teams that rely on daily monitoring with internal RUM tools.

In what cases can this measurement become misleading?

For sites with pronounced seasonal traffic, CrUX may aggregate very different load periods. An e-commerce site that spikes in December and plods along in January will have its CrUX data for February mix high and low season periods. The result: an average score that does not correspond to any operational reality.

Finally, low-traffic sites remain in the dark. Not enough CrUX data? PSI shows only lab data, but Google does not explicitly state whether the absence of CrUX affects ranking. The most likely assumption is that the Core Web Vitals signal is neutral (neither a bonus nor a penalty) in this case. But this is a gray area that deserves clarification.