December 20, 2025

Google PageSpeed Insights: What the Scores Mean and How to Improve Them

Google PageSpeed Insights is often treated as the simplest way to test website speed: enter a URL, run the analysis and receive a score from 0 to 100. Then the questions begin. One website scores 97 while another gets 72. The mobile version may show 58 while desktop reports 96. Running the same test again can produce a slightly different result. This makes it easy to draw the wrong conclusion: that a website with a score of 100 is automatically fast and SEO-friendly while anything around 70 needs to be completely rebuilt. PageSpeed Insights is much more useful when treated not as an exam, but as a diagnostic tool. It can help identify performance problems, reveal potential bottlenecks and determine what should be investigated first.

What is Google PageSpeed Insights?

PageSpeed Insights is a Google tool for analysing the performance of web pages. It combines automated testing with information about real-user experience when enough real-world data is available for a particular page or site. That is why the report should not be reduced to the large Performance number at the top. Context matters:

whether you are reviewing mobile or desktop;
whether real-user data is available;
which metrics are causing problems;
what the diagnostics identify;
whether the issue is noticeable to actual visitors.

What does the score from 0 to 100 mean?

Lighthouse broadly groups its performance scores into three ranges:

0–49 — poor;
50–89 — needs improvement;
90–100 — good.

These ranges are useful guidance, but they are not an absolute assessment of the quality of a business website. The difference between a score of 45 and 85 can be significant. The practical difference between 96 and 100 may be much smaller. As a score approaches 100, earning the final few points may require increasingly more development effort for progressively less noticeable benefit.

Does a website need a perfect 100/100 PageSpeed score?

No. A perfect score can be a satisfying technical result, but it should not become the primary objective of a website project. Consider two websites. The first scores 100 but has confusing navigation, weak content and a form that visitors do not want to complete. The second scores 93 but presents its key information quickly, works comfortably on mobile devices and makes the desired action easy to complete. From a business perspective, the second website may be significantly stronger. Performance optimization should improve the user experience, not simply improve a test score.

Why does the PageSpeed score change?

It is common to run the same test several times and receive slightly different results. This is normal. Laboratory testing can be affected by:

network conditions;
server load;
third-party scripts;
advertising and analytics services;
dynamic content;
caching;
JavaScript execution timing;
other conditions during the individual test run.

A website should therefore not be judged by one random test. If the result moves between 88 and 94, that does not mean the website changes from “bad” to “good” every few minutes. It is more useful to identify issues that appear consistently.

Why is the mobile score often lower than desktop?

It is very common to see a website score around 95 on desktop and perhaps 60–70 on mobile. This does not necessarily indicate an error. Mobile testing represents more constrained conditions, making expensive JavaScript, oversized images and third-party services more visible. The mobile report can therefore reveal performance problems that are difficult to notice on a powerful development machine. Optimizing only the desktop version to produce an attractive score is rarely a useful strategy.

PageSpeed Insights and Core Web Vitals are not the same thing

These concepts are frequently confused. PageSpeed Insights is a testing and diagnostic tool. Core Web Vitals are user-experience metrics related to loading performance, responsiveness and visual stability. PageSpeed Insights can display information related to Core Web Vitals, but the overall Performance Score is not itself a Core Web Vitals score. A Lighthouse Performance score of 100 therefore does not automatically mean every real visitor receives a perfect experience.

Lab data vs real-user data

One of the most important distinctions in PageSpeed Insights is the difference between laboratory testing and real-user data.

Lab data

Lab data comes from a controlled test under predefined conditions. It is particularly useful during development because a team can identify a technical problem, make a change and repeat the test.

Field data

When enough information is available, PageSpeed Insights can show aggregated data from real Chrome users. This reflects how the website has performed in actual conditions across different devices, connections and user environments. The two types of data can differ, and that is not necessarily a contradiction. A laboratory test tells you what happened during a particular simulated run. Field data helps you understand what real users experienced over a period of time.

What commonly makes websites slow?

1. Oversized images

A photograph directly from a camera may be several megabytes even though the website only needs a much smaller asset. Large hero images are a particularly common source of unnecessary page weight. Correct dimensions, compression and appropriate modern image formats can make a substantial difference.

2. Excessive JavaScript

JavaScript enables interactive interfaces, but the browser still needs to download, parse and execute the code. The more unnecessary logic a page loads, the greater the potential cost, particularly on mobile devices. The issue is rarely that JavaScript exists. The issue is how much is being delivered and when it needs to execute.

3. Third-party scripts

Analytics, chat tools, maps, advertising platforms, personalization systems and other widgets can add network requests and JavaScript. Each service may provide genuine business value. However, when many third-party services run on the same page, their combined performance cost can become significant.

4. Web fonts

Multiple font families, weights and files can also increase the amount of data a visitor needs to download. Complex design systems should periodically review whether every included font variation is actually being used.

5. Video and sophisticated animation

Video backgrounds, WebGL, 3D and advanced motion design can be appropriate for a particular product or brand. They should, however, be designed with performance in mind, particularly for mobile users.

6. Slow server response

Not every performance problem exists in the browser. If the server takes too long to generate a response, optimizing images alone will not solve the issue. The bottleneck may involve hosting, databases, server-side logic, caching or external APIs.

What should you fix first in PageSpeed Insights?

Not every warning deserves the same priority. A useful optimization process usually starts with problems that:

affect primary page content;
occur across many pages;
are particularly noticeable on mobile devices;
affect real Core Web Vitals;
can be fixed without damaging important business functionality.

Optimizing a massive hero image may provide much more practical value than spending hours removing a tiny amount of CSS purely to gain one additional Lighthouse point.

Should you remove analytics to improve PageSpeed?

Not automatically. Analytics platforms, CRMs, advertising pixels and other services can affect performance, but they may also perform critical business functions. A better approach is to ask:

Are all of these services still being used?
Are any tools duplicated?
When do the scripts load?
Can some of them be delayed?
What business value does each service provide?

Performance is an important website characteristic, but it is not the only one.

Does PageSpeed affect SEO?

Two different ideas need to be separated here. Google uses Core Web Vitals within its ranking systems and recommends providing a strong overall page experience. However, a PageSpeed score of 90, 95 or 100 is not a direct formula for a particular position in search results. A page scoring 100/100 does not automatically outrank more useful and relevant content with a lower Lighthouse score. SEO depends on many additional factors, including content quality, search intent, site architecture, internal linking, indexing and other signals. Website performance should therefore be improved primarily because it creates a better website, not because a perfect test score provides a secret ranking advantage.

Does website speed affect conversion?

A slow or unstable page can create unnecessary friction in a customer journey. However, it would be misleading to promise a universal increase in sales after improving a PageSpeed score by a particular number of points. Performance may be a major problem on one website. On another, the main conversion issue may be pricing, the offer, trust, the form, traffic quality or page structure. After performance work, businesses should review real business metrics in addition to technical scores.

How should you test a website after optimization?

After implementing changes, use several checks rather than relying on a single number:

run PageSpeed Insights more than once;
test mobile and desktop separately;
use the website on a real smartphone;
check important internal pages, not just the homepage;
review Core Web Vitals in Search Console where sufficient data exists;
confirm that optimization did not break functionality.

The final point is particularly important. A score of 100 provides little value if a contact form no longer works, analytics disappears or important content is rendered incorrectly.

When does a website genuinely need performance optimization?

Performance deserves attention when:

visitors visibly wait for important content;
mobile performance is substantially worse than desktop;
the interface feels unresponsive;
page elements move unexpectedly during loading;
Core Web Vitals consistently indicate problems;
important pages are overloaded with resources;
new third-party services have made the website noticeably heavier.

On the other hand, spending development budget purely because a score of 97 feels psychologically unsatisfying and someone wants to see 100 may not be the best investment.

The TimeKairos approach

At TimeKairos, we do not treat a 100/100 PageSpeed score as an objective in itself. Performance is one part of overall website quality alongside UX, design, SEO, accessibility and business functionality. The first step is to identify the real problem. The second is to understand its cause. Only then should a team decide which changes will provide the greatest benefit. Sometimes optimizing images and third-party scripts is enough. In other projects, the bottleneck may be architectural, server-related or the result of accumulated technical debt. A useful performance audit should therefore end not with dozens of alarming red warnings, but with clear priorities: what is critical, what is worth improving and what can reasonably remain unchanged.

Conclusion

Google PageSpeed Insights is a useful tool, but a single number cannot describe the overall quality of a website. A score between 90 and 100 is a useful target. However, 100/100 does not guarantee high Google rankings, conversions or perfect UX. What matters more is understanding what slows the page down, how the website performs for real users and whether technical problems create friction during important actions. The purpose of performance optimization is not a perfect PageSpeed screenshot. It is a website that feels fast, stable and comfortable for the people who actually use it.