21.07.2026
Technologies & CMS

A score of 100 on PageSpeed Insights guarantees neither a Google position nor conversions. This guide explains how to read lab and real-world data, master the Core Web Vitals and prioritise useful optimisations.

Getting a score of 100 on Google PageSpeed Insights means that a page reaches the maximum performance score calculated by Lighthouse under given test conditions. It is an excellent technical result, but it guarantees neither instant loading for all users, nor a better position in Google, nor an increase in conversions.
For a professional site, the objective must be broader: display useful information quickly, respond to interactions without delay and avoid unexpected interface shifts.
Google PageSpeed Insights helps identify the main obstacles to this experience. The tool combines lab tests and, when the data is available, measurements from real Chrome users.
So how do you improve your PageSpeed score? What do the LCP, INP and CLS indicators correspond to? And do you really need to reach 100/100 on mobile and desktop?
ID Interactive is a digital agency based in Arradon, near Vannes (Morbihan), since 2006, and a Google Partner, Meta Business Partner and Shopify Partner. It designs websites and web applications, building performance into their development from the start.
This guide explains how to interpret PageSpeed Insights and prioritise the optimisations that bring real improvement.
Google PageSpeed Insights is a free tool for analysing a web page's performance.
It is available at pagespeed.web.dev.
To run a diagnosis, simply enter the URL of the page you want to test.
The tool then provides information on performance in mobile and desktop environments.
It presents in particular:
PageSpeed Insights therefore helps guide an improvement process.
However, it replaces neither a full performance audit nor user testing.
The PageSpeed Insights performance score is calculated from several metrics collected during a Lighthouse analysis.
It ranges from 0 to 100.
Three categories make it easy to read the result:
A score of 100 corresponds to the maximum of this scale for the test carried out.
It does not mean the site is running at 100% of its possible speed or that all its visitors enjoy identical conditions.
No. Google does not ask websites to achieve a perfect score.
A score of 90 or more is already considered good on the Lighthouse scale.
It may be worth pursuing some optimisations beyond this level, particularly when they genuinely improve loading or interactivity.
On the other hand, trying to gain the last few points at the cost of many technical changes may be of limited value.
A result must be assessed in light of usage, the site's constraints and real user data.
Even if a page has not been modified, its score can vary.
Various parameters can come into play:
It is therefore recommended to run several comparable tests before concluding that a change has improved or degraded performance.
No.
PageSpeed Insights analyses mobile and desktop environments separately.
A page can get an excellent score on desktop and have difficulties on mobile.
Simulation conditions, hardware constraints and content organisation can explain these differences.
The diagnosis must therefore take both environments into account, with particular attention to the devices actually used by the site's audience.
This is the most important distinction for correctly interpreting PageSpeed Insights.
The tool can display two sets of information that do not measure exactly the same thing.
Lighthouse runs an automated test in a simulated environment.
This diagnosis makes it possible to examine how a page loads and identify technical points to fix.
It provides in particular indicators such as:
The performance score out of 100 comes from these lab measurements.
These tests are particularly useful for diagnosing problems and comparing optimisations carried out under similar conditions.
PageSpeed Insights can also present measurements collected from real Chrome users.
This data comes from the Chrome User Experience Report, or CrUX.
It covers a rolling 28-day period and is aggregated to preserve anonymity.
It makes it possible in particular to study the loading performance, responsiveness and stability of the pages actually visited.
Results are generally presented at the 75th percentile.
This means that a reference value gives a picture of the experience of a large majority of visits, without hiding the difficulties encountered by some users.
A page can get 100/100 in a Lighthouse test and have insufficient Core Web Vitals in real conditions.
This difference can be explained by the diversity of devices, connections, browsers and behaviours.
It may also come from loading third-party services or interactions that are not reproduced during the automated test.
Conversely, a page with good real-world data may occasionally get a lab score below 90.
To manage the performance of a site in production, you therefore need to look at both types of measurement.
Some pages do not generate enough CrUX data.
PageSpeed Insights may then display data aggregated at domain level, or offer no usable real-world measurements.
The absence of CrUX data does not mean the page is slow.
It simply indicates that the available information does not allow the measurement to be established under the conditions set by the tool.


The Core Web Vitals make it possible to assess three aspects of the browsing experience: loading, interactivity and visual stability.
In 2026, the main metrics are LCP, INP and CLS.
These thresholds apply to the analysis of user experiences, in particular at the 75th percentile of measurements.
LCP (Largest Contentful Paint) measures the time needed to display the largest content element visible in the browser window.
This may be a main image, a block of text or another element taken into account by this metric.
A high LCP can have several causes:
LCP therefore does not measure server speed alone.
It represents a significant part of the loading perceived by the user.
INP (Interaction to Next Paint) measures the page's responsiveness to user interactions.
It takes into account the delay between an interaction and the corresponding visual update.
For example, if a user clicks on a menu and waits a long time before it appears, the experience may be degraded.
INP problems are often linked to heavy JavaScript processing or tasks that occupy the browser's main thread for too long.
INP replaced FID (First Input Delay) among the Core Web Vitals in March 2024.
Note: the TBT shown by Lighthouse is a useful lab indicator for diagnosing main thread blocking, but it is not a direct measure of the real INP of all interactions.
CLS (Cumulative Layout Shift) measures unexpected shifts of certain visible elements.
These movements can occur when an image, an ad, a font or a component loads after display has started.
For example, a user is about to click a button. An image appears above it, shifts the content and suddenly moves the button.
Good design must reserve the space needed for elements and avoid these unexpected changes.
CLS is a unitless value.

Improving performance relies on technical actions suited to the problems detected.
There is no single setting that takes every site to 100/100.
Here are the most useful optimisations to look at.
Images can account for a large share of the resources downloaded.
To limit their impact:
Images below the fold can be lazy-loaded where relevant.
On the other hand, avoid lazy-loading the main image responsible for the LCP.
This resource must generally be identified and loaded early enough.
Scripts can slow down display or delay interactions.
This applies both to features developed for the site and to third-party tools: analytics, advertising, chat, widgets or animations.
Several approaches can be considered:
The aim is not to remove all JavaScript.
It is about limiting processing that brings no value to the user.
Large or poorly organised style sheets can delay display.
Web fonts can also cause slowdowns or layout changes.
It is useful to limit unused resources, choose appropriate file formats and adapt loading strategies.
These optimisations must be assessed according to real graphic needs.
The server's initial response time can affect how content loads.
Several elements need to be examined:
Using a CDN can be relevant for some projects, but it does not automatically fix an inefficient architecture or server processing.
Caching makes it possible to reuse certain resources without having to recalculate or fully download them on every visit.
It can apply to static files, server responses or certain intermediate content.
A suitable configuration must take into account how often content is updated and confidentiality needs.
Poorly managed caching can also display an outdated version of a page after publication.
You therefore need a consistent strategy for retention times and invalidation.
Tracking tools, advertising modules and external components sometimes add JavaScript, network connections and extra processing.
Before integrating a new tool, you need to examine its real usefulness.
For each service, several questions can be asked:
This work must preserve legal obligations, measurement needs and essential features.
Animations can make certain interactions easier to understand.
But excessive or poorly designed effects can cause slowdowns, disrupt reading and make the site harder to use.
Animations must have a clear purpose and respect users' preferences when they are expressed.
A new feature, a marketing tool or a graphic change can affect performance.
It is therefore best to check important pages regularly.
Monitoring must cover in particular home pages, service pages, product pages, booking journeys and critical commercial features.
The PageSpeed Insights diagnosis can be translated into an action plan.
Recommendations must be adapted to the site tested.
A full diagnosis generally starts with the most serious problems on the pages that matter most for users and the business.
A simple method gives usable results.
Testing only the home page is not enough.
A professional site may include different templates with very different technical behaviour.
Select in particular a sales page, an article, a contact page and, depending on the site, a product page or a booking journey.
Open PageSpeed Insights and enter the URL.
Examine the two environments separately.
Keep the main results so you can compare them later.
When CrUX data is available, start by looking at the LCP, INP and CLS values.
Identify the insufficient metrics.
Also check whether the data concerns the exact URL tested or a wider set of pages on the domain.
The lab report helps you look for possible technical causes.
For example, a high LCP may lead you to analyse the main image, blocking resources or server response time.
Prioritise interventions according to:
A problem present on all product pages may be more important than a detail affecting a single rarely viewed page.
After an optimisation, run several lab tests again under comparable conditions.
For real user data, take into account its rolling 28-day collection period.
A change in the Lighthouse score can be visible immediately, while its effect in CrUX measurements appears gradually.
No. A Lighthouse score of 100/100 is not a direct ranking factor equivalent to first position on Google.
Google uses the Core Web Vitals among the signals taken into account by its ranking systems.
But SEO relies on many other elements: content relevance, indexability, architecture, links, quality of answers and search context.
A page with a lower PageSpeed score can appear ahead of a page scoring 100/100 if it better meets the internet user's intent.
Lighthouse also offers a category of SEO checks.
It examines certain technical fundamentals that can be automated.
However, even a score of 100 in this category does not mean the site has a complete SEO strategy.
The tool does not replace an analysis of keywords, competitors, editorial quality or conversions.
ID Interactive combines these different dimensions in its services for SEO.
Loading problems can make navigation harder and lead some internet users to abandon a visit.
But you should not claim that improving the PageSpeed score automatically reduces bounce rate.
In Google Analytics 4, this indicator depends in particular on duration and engagement criteria.
The guide to bounce rate in GA4 explains how to calculate and interpret it.
A technically accessible, high-performing site is a useful foundation for visitors and crawling systems.
But a score of 100 on PageSpeed Insights does not guarantee that a page will be cited in AI assistants' answers.
Google does not require any specific PageSpeed score to appear in its AI Overviews or AI Mode.
GEO visibility also depends on the quality of information, the relevance of sources and the ability of systems to access the content.
It is therefore better to treat performance as a component of site quality, not as an AI citation technique.
Technical optimisation can help reduce some of the resources needed to run a site.
A better-sized image, a removed script or an avoided request can limit the amount of data transferred and some processing.
These improvements can be valuable both for performance and for a digital sobriety approach.
No.
Loading speed is not a direct measure of environmental impact.
The impact of a digital service depends in particular on its infrastructure, the equipment used, hardware lifespan, processing and usage.
Reducing page weight can be relevant, but it does not on its own allow a precise calculation of a reduction in CO₂ emissions.
An eco-design approach can start by thinking about real needs.
This means in particular:
Tools such as EcoIndex can provide additional indicators on certain aspects of a page's design and weight.
However, their estimates must be distinguished from a full environmental analysis.
ID Interactive presents these issues in its approach to responsible digital eco-design.


The choice of technologies can make certain optimisations easier.
But no technology alone guarantees a score of 100 on PageSpeed Insights.
The quality of the result also depends on the architecture, the content, the components developed, the hosting and third-party resources.
Next.js is a React-based framework used to develop websites and web applications.
It offers several mechanisms useful for optimising performance:
Rendering suited to the project. Depending on the architecture chosen, pages can be generated in advance, rendered server-side or use different caching strategies.
Image optimisation. The framework's components and functions make it possible to adapt resources to devices and display needs.
Font management. Dedicated features make fonts easier to load and optimise.
Control of the JavaScript sent to the browser. With a suitable architecture and components, it is possible to limit the code executed client-side.
These capabilities require skilled implementation.
A Next.js site can remain slow if its pages include overly heavy scripts, poorly loaded images or costly server calls.
Discover ID Interactive's expertise as a Next.js agency.
Payload CMS is a headless content management system that lets you manage information independently of how it is presented on the site.
This architecture can make it easier to design interfaces suited to the project's needs.
Content can be made available to applications via APIs and integrated into different environments.
It can also make it possible to control the data sent to interfaces.
However, using a headless CMS does not automatically reduce the number of requests or the weight of each page.
These results depend on integration choices, the APIs used, caches and presentation components.
ID Interactive develops projects using this architecture and presents its services as a Payload CMS agency.
Combining these two tools separates editorial management from the presentation layer.
It can suit professional sites requiring a customised interface, structured content and specific features.
This separation makes some development choices easier, but it also requires a suitable technical organisation.
Performance must be checked on the delivered site, under production conditions.
Screenshots illustrating the performance of the agency's projects may show the results obtained during specific tests.
These results are useful for understanding the optimisation possibilities of a project.
However, they must be accompanied by the information needed to interpret them:
A one-off score must not be presented as a guarantee of permanent performance for an entire site.


It is generally better to take performance into account from the design stage rather than multiplying fixes after going live.
This approach starts with defining the need.
A showcase site, an e-commerce catalogue or a business application do not require the same features or the same technical constraints.
The specifications can set out:
This analysis makes it possible to choose an architecture proportionate to its use.
Rather than contracting only for a Lighthouse score of 100, it can be more useful to set objectives relating to the Core Web Vitals, feature responsiveness and priority pages.
The measurement method and conditions must also be specified.
Performance must be monitored after going into production, since the site's usage and content evolve.
A technically fast site can remain difficult to use.
Clear menus, quality information and easy forms are also essential.
The aim is to design a service that properly meets its users' needs.
ID Interactive builds this approach into its website creation services, with an architecture and features suited to the project.
ID Interactive is a digital agency based near Vannes (Morbihan) since 2006.
The agency supports companies in creating custom websites and web applications, combining design, development and technical optimisation.
A performance improvement approach can include:
1. Initial diagnosis. Identify the important pages and examine PageSpeed Insights, Lighthouse and Core Web Vitals data.
2. Analysis of causes. Look for the resources, scripts, processing or configurations responsible for slowdowns.
3. Prioritisation. Select fixes according to their expected impact and the site's needs.
4. Implementation. Optimise media loading, code, caching or architecture where necessary.
5. Acceptance testing. Check that changes preserve features, accessibility and content.
6. Follow-up. Monitor real-world performance after publication and as the site evolves.
The aim is not only to improve a technical indicator, but to offer a reliable, fast experience suited to users.
To discuss a creation or redesign project, you can request a quote from ID Interactive.
A score of 100 on PageSpeed Insights is an interesting technical result, but it should not become an end in itself.
A score of 90 or more is already considered good by Lighthouse.
The priority is to check that the Core Web Vitals are satisfactory for real users and that the site's features remain pleasant to use.
Optimisations must be assessed in terms of their concrete impact: content loading, responsiveness, stability, journey quality and continuity of services.
Finally, the PageSpeed score does not replace an SEO audit, a full accessibility test or an environmental analysis.
A high-performing site is above all a site that responds quickly and correctly to its users' needs, in real browsing conditions.
Article originally published on 6 July 2023.
Yes, it is a free Google tool that analyses a page's performance from its URL, on mobile and desktop. It combines Lighthouse lab tests and, when there is enough volume, measurements from real Chrome users. It replaces neither a full performance audit nor user testing.
INP (Interaction to Next Paint) measures page responsiveness by taking into account the delay between an interaction and the corresponding visual update. It replaced FID (First Input Delay) among the Core Web Vitals in March 2024. An INP of 200 milliseconds or less is considered good.
No. The TBT (Total Blocking Time) shown by Lighthouse is a lab metric that is useful for diagnosing main thread blocking, but it is not a direct measure of the real INP of all interactions. INP is part of the Core Web Vitals, assessed on real user experiences.
No. Images below the fold can be lazy-loaded where relevant. The main image responsible for LCP, however, should avoid lazy loading: it should generally be identified and loaded early enough.
The Lighthouse score can change immediately: it is recommended to rerun several lab tests in comparable conditions. Real user data (CrUX), on the other hand, covers a rolling 28-day period: the effect of a change appears there gradually.