European web performance in 2026: does the north-south divide really exist?

Core Web Vitals, TTFB, mobile, compression: a state of the art on performance indicators before the Orilyt 2026 barometer publishes on June 16.

Since the generalization of Core Web Vitals as a Google ranking signal in June 2021, web performance has left the purely technical terrain to become a commercial and strategic topic. Five years later, where do European sites really stand against these now-standard requirements?

Publicly available global figures, from the Chrome User Experience Report and HTTP Archive, give a global view but aggregate very heterogeneous ecosystems. The legitimate question becomes one of national specificities: does a German site have better LCP than a Spanish site? Does French mobile catch up with Swiss desktop?

This article takes stock of major web performance indicators worldwide in 2026, from widely recognized public sources. It sets the context before the publication, on June 16, 2026, of the Orilyt barometer that will measure these same criteria across five distinct European panels.

Why performance remains a burning topic in 2026

Five years after Core Web Vitals were integrated into Google ranking, performance is no longer optional. Three concurrent factors keep it at the heart of technical and commercial debate.

Continuous standard evolution: INP replaced FID in 2024

In March 2024, Google replaced First Input Delay (FID) with Interaction to Next Paint (INP) in Core Web Vitals. This change is not cosmetic: INP measures responsiveness across all user interactions, not just the first. The "good" threshold is set at 200 milliseconds, and many sites previously green on FID find themselves orange or red on INP.

This evolution reflects increased demands: it is no longer enough to display quickly, you must also respond quickly to clicks, taps, and inputs. For sites with significant third-party JavaScript (analytics, chat, tag manager), it is a paradigm shift that reshuffles the cards of performance audits.

Agencies and freelancers who based their arguments on FID in 2022-2023 must now adjust their discourse and deliverables. INP has become in 18 months the most discriminating metric of technical audits.

What public global data measures

The Chrome User Experience Report (CrUX), fed by Chrome users who accept data sharing, is the official reference for measuring field performance. In Google's 2024 publications, about two thirds of global origins passed the "good" thresholds on combined LCP, INP, and CLS, in continuous progression since 2021.

The HTTP Archive Web Almanac, which crawls over 16 million sites annually, gives a complementary view. Its 2024 Performance chapter confirms the slow progression of the web fleet on Core Web Vitals, while pointing to a persistent gap between top sites and the long tail of SMBs or nonprofits.

These global figures nevertheless aggregate very heterogeneous ecosystems: heavy American e-commerce sites, European WordPress blogs, Asian institutional sites. Drawing operational conclusions for a French or German agency from global aggregates remains unsatisfactory.

Why a European focus remains missing

Existing barometers rarely compare technical practices between European countries on a broad sample. When national studies appear, they are often commercial (funded by a host or tool) and difficult to cross-reference due to divergent methodologies.

This is precisely the gap the Orilyt barometer, published on June 16, 2026, intends to fill for performance as well as security. Five independent national panels were scanned with the same technical grid, allowing for the first time a direct comparison between France, Germany, Spain, Belgium, and Switzerland.

Pending this publication, let us look at what global public data already tells us about the major performance axes, and where gaps between national ecosystems are most likely.

Core Web Vitals: where do sites really stand in 2026?

LCP, INP, CLS: three metrics alone decide the quality perceived by the user. Their global progression is slow but real, with strong disparities depending on site profiles.

LCP: the heavyweight of perceived performance

Largest Contentful Paint measures the render time of the largest visible element above the fold, often a hero image or text block. The "good" threshold is set at 2.5 seconds on mobile. Above 4 seconds, LCP falls into "poor" and penalizes SEO ranking.

According to CrUX reports aggregated by Google, about 70% of origins passed the "good" threshold on LCP in 2024, up from 2021. This progression reflects the generalization of native lazy loading, massive conversion to WebP and AVIF, and improved hosting infrastructure.

To go further on this metric and its optimization levers, see our complete guide to improving LCP.

CLS: the best-mastered metric

Cumulative Layout Shift measures visual shifts during loading. It is the metric on which the global web fleet has progressed most. Over 80% of origins passed the "good" threshold (below 0.1) in 2024, according to Google's reports.

This mastery is explained by the relative simplicity of fixes: declaring image dimensions, reserving space for ad banners, using font-display: swap with preloading. Three practices widely adopted since 2021.

The risk today lies in unmastered additions: a new poorly integrated cookie banner, a CRM integration that pushes content, can degrade the CLS of a previously well-rated site. Continuous monitoring has become the only way to detect these regressions.

INP: the challenger that reshuffles the cards

Interaction to Next Paint measures responsiveness to user interactions. The "good" threshold is 200 milliseconds. Many poorly optimized WordPress sites sit between 250 and 500 ms, mainly due to third-party scripts loaded on the main thread.

According to Google, the rate of origins passing "good" on INP is lower than that of LCP and CLS. This metric has become in 18 months the most discriminating for identifying sites overloaded with marketing JavaScript.

For agencies, INP is also a concrete commercial argument. An audit mission revealing a red INP, followed by an optimization of third-party scripts bringing the metric back to green, tells a quantified story presentable to a non-technical decision-maker.

Server performance: TTFB and infrastructure

Beyond visible Core Web Vitals, server performance conditions everything else. TTFB, HTTP/2-3 adoption, and Brotli compression remain the most powerful levers on the infrastructure side.

TTFB: a often-ignored bottleneck

Time to First Byte measures how long the server takes to respond. Above 800 milliseconds, it mechanically dooms LCP to be poor, regardless of front-end quality. Yet it is the least surveyed metric by free consumer audits.

Variations are enormous depending on the host. On a low-end shared host shared with hundreds of other sites, TTFB can exceed 1.2 seconds. On a properly configured VPS or specialized managed WordPress hosting, it drops below 200 ms.

This disparity explains the importance of a dedicated TTFB analysis from the diagnostic phase, before any front-end recommendation.

HTTP/2 and HTTP/3: real adoption rate

HTTP/2, generalized since 2018, enables request multiplexing, meaning sending multiple files simultaneously over a single connection. HTTP/3, based on QUIC, adds resilience to packet loss. Both protocols significantly reduce loading time, especially on mobile connections.

According to W3Techs and HTTP Archive, HTTP/2 is now present on over 80% of major sites, but HTTP/3 remains below 30%. Many sites still run on HTTP/1.1 without anyone noticing, due to lack of dedicated audit.

Verifying the protocol in use takes seconds, but it is one of those signals that free consumer audits do not always report. On institutional sites or barely-monitored SMBs, it is typically the quick win that an Orilyt audit immediately highlights.

Brotli versus Gzip compression

Brotli, the compression format developed by Google and published in 2015, offers significantly higher weight reduction than Gzip on text files. It is supported by all modern browsers since 2018 and activatable in a few minutes on most hosts.

Despite this simplicity, its adoption remains partial. HTTP Archive observed in 2024 that a substantial share of sites continue to use Gzip alone or no compression on certain resources. An audit must therefore systematically verify not only the presence of compression but also its level (Brotli levels 4 to 11 depending on file types).

On unoptimized WordPress sites, enabling Brotli combined with a modern cache plugin (WP Rocket, LiteSpeed Cache) remains one of the best impact-to-effort levers. It is also one of the 12 criteria that will be measured panel by panel in the June 16 barometer.

Mobile: the terrain where everything plays out

Google has indexed mobile-first since 2020. Yet the gap between mobile and desktop performance remains systematic and largely underestimated by site editors.

Mobile-first since 2020, but the gap persists

Since Google generalized mobile-first indexing in September 2020, it is the mobile version that is used by default for ranking. Yet on most professional sites, the Lighthouse mobile score remains 20 to 40 points lower than the desktop score for the same resources.

This dichotomy is mechanically explained: the Lighthouse mobile simulation applies a degraded 4G connection and a throttled processor, conditions closer to the reality of a rural user with a mid-range mobile. These conditions are sometimes ignored by teams working on large fiber screens.

For a complete method covering mobile-specific levers, see our guide to reducing load time, which details the optimizations to apply in mobile priority.

Page weight in continuous increase

HTTP Archive measures every year the median weight of a mobile page loaded. The curve has climbed year after year since 2017, exceeding 2 MB in 2024 according to published figures. This inflation is largely driven by unoptimized images, autoplay videos, and third-party JavaScript.

For freelancers and agencies auditing sites daily, this trend changes the nature of work. Optimizing a site is no longer just improving a threshold, it is constantly fighting against a structural drift. Continuous monitoring becomes a practical condition of maintenance.

The June 16 barometer will notably measure median page weight by national panel. The expected disparities between ecosystems (e-commerce, showcase, e-administration) are one of the interests of the comparison.

Images, leading cause of weight

Images account for nearly half the median weight of a web page according to the HTTP Archive Almanac 2024. Compression, conversion to WebP or AVIF, responsive sizing via srcset, and lazy loading are the four levers that solve most of the problem.

On WordPress, plugins like ShortPixel, Imagify, or EWWW automate the complete chain. But they do not cover everything: the hero image visible on arrival requires specific treatment with fetchpriority="high" and possible preloading.

This nuance is often neglected and costs several tenths of a second on LCP. It is typically the quick lever that an Orilyt audit immediately highlights, and that turns a simple optimization mission into a discussion on the global loading strategy.

European national ecosystems: indicators already available

Without a dedicated barometer, several indirect signals suggest that European national web ecosystems present different performance profiles. Here are three hypotheses largely verifiable before the June 16 publication.

Variations by dominant host

Each national market has its dominant hosts. OVH and Infomaniak in France and Switzerland, Hetzner and IONOS in Germany, Strato and Telekom in DACH, Combell in Belgium. These hosting ecosystems present different infrastructure qualities and default configurations.

This diversity directly conditions average national TTFB. A fleet dominated by a host that supports HTTP/3 and Brotli level 11 by default mechanically displays a better global score than that of a fleet still on HTTP/1.1.

The June 16 barometer will not directly measure the host, but the consequences (TTFB, protocol, compression) remain observable and will reveal these structural differences between countries.

Variations by dominant CMS

WordPress dominates everywhere in Europe but with varying intensities by country. The share of WordPress sites on a German sample may differ noticeably from that of a Belgian or Swiss sample. These variations condition average performance profiles, because unoptimized WordPress statistically weighs more than other CMSs.

Other CMSs play a local role: Typo3 and Contao remain very present in the Germanic ecosystem, PrestaShop is strong in France and Belgium, Drupal in France for the public sector. Each brings its own average performance profiles.

The June 16 barometer will allow crossing performance indicators with observable CMS signatures for each panel, revealing for the first time the crossed impact on national performance.

Mobile versus fixed audience pressure

Mobile penetration rate and average network quality vary significantly between European countries. According to European Commission DESI reports, 5G and fiber deployment is not uniform between Switzerland, Germany, France, Belgium, and Spain.

These access infrastructure variations translate into perceived performance. A site that loads in 3 seconds on a German 5G connection at 200 Mbps does not have the same user-side perception as a 3-second load on a throttled Spanish 4G.

The Orilyt barometer does not measure end-of-chain access infrastructure, but the technical quality of the site itself. The CrUX field performance will integrate these factors and remain the ultimate barometer on the user experience side.

What the Orilyt barometer of June 16, 2026 will bring

Global figures provide a frame. For European agencies auditing sites daily, the real question is: how do European sites position, country by country, on these performance indicators?

Five independent national panels scanned

The Orilyt 2026 barometer scanned five independent national panels drawn from public sectoral databases: .fr (France), .de (Germany), .es (Spain), .be (Belgium), and .ch (Switzerland). Each panel was constituted to reflect the real diversity of the national web fabric, not just the top of the most-visited sites.

This multi-country approach on representative panels is what was missing from the landscape of existing barometers, either focused on the world top or limited to a single geography. The direct comparison between five European ecosystems opens unprecedented questions about the supposed north-south divide of web performance.

Four performance criteria among the twelve

Of the 12 homogeneous technical criteria of the barometer, four concern performance directly: presence and level of compression (Gzip/Brotli), mobile-friendly compliance, presence of meta description (SEO signal but also technical maturity indicator), and viewport declaration adapted to mobile.

To these four criteria are added indirectly other indicators linked to perceived performance, like WordPress version exposure (which can reveal unmaintained sites with degraded performance). The homogeneous grid allows direct comparison country by country.

Why it matters for agencies

For a freelancer or agency auditing client sites, knowing the average state of the national market is a concrete discussion argument. Being able to say "your site is in the lower third of the French market on compression" makes the conversation tangible, where "your site is not optimal" remains vague.

The Orilyt barometer will be published in open access on June 16, 2026, with a press kit, detailed graphics, and an interactive page allowing positioning of a given site against its national panel average.

Pending this publication, launching an Orilyt audit on a site provides immediate access to a complete diagnosis on the twelve barometer criteria, plus over forty other universal tests and over forty-five CMS-specific tests.

The state of web performance in 2026 remains a moving landscape. Core Web Vitals progress slowly but really at the global scale, driven by the generalization of front-end best practices and the improvement of hosting infrastructure.

Nevertheless, the gap between global top 1 000 and SMB long tail remains massive, and mobile performance remains systematically behind desktop. The continuous increase in mobile page weight reflects a structural drift that only agencies engaged in continuous maintenance manage to contain on their portfolios.

The Orilyt barometer published on June 16, 2026 will bring, for the first time, a rigorous comparison between five European national ecosystems on these criteria. In the meantime, launching a free audit on your site or a client's provides immediate access to an objective measurement of its position on the twelve criteria that structure the analysis.

Your most frequent questions

Sources and references

• Google web.dev, Core Web Vitals — official reference for LCP, INP, CLS thresholds
• Chrome User Experience Report (CrUX) — Google documentation on field performance data
• HTTP Archive, Web Almanac 2024 Performance chapter — annual state of global web performance
• HTTP Archive, Web Almanac 2024 Page Weight — evolution of web page weight and image share
• W3Techs, Web servers and HTTP versions usage statistics — HTTP/2 and HTTP/3 adoption
• European Commission, DESI Digital Economy and Society Index — national infrastructure indicators