A free mockup of your website, with no commitment. Learn more
All articles
Tech August 2026 · 9 min

Lighthouse score: how to get above 95 on mobile in 2026

Lighthouse is the 2026 standard for measuring a website's technical quality. Reaching 95+ on mobile now separates the sites that rank from the sites that struggle.

Contents

What Lighthouse measures in 2026

Lighthouse is Google’s official tool for auditing a website’s technical quality. It measures 4 dimensions:

  1. Performance: loading speed, smoothness, responsiveness (LCP, INP, CLS, FCP, TBT)
  2. Accessibility: WCAG compliance, readability, keyboard navigation
  3. Best Practices: security (HTTPS), optimised images, console errors
  4. SEO: metadata, indexability, basic structured data

Each dimension is scored out of 100. The overall score is the weighted average.

Google’s thresholds:

  • 0-49: red (poor)
  • 50-89: orange (needs improvement)
  • 90-100: green (good)

But in 2026, on competitive searches, 90 is no longer enough: you need to aim for 95+ on all 4 dimensions to beat the competitors who have done the work. This article explains how.

Why 95+ and not 90+

Three reasons to push beyond 90.

Core Web Vitals weighting. Since the March 2026 Core Update, the Core Web Vitals (LCP, INP, CLS) carry more weight in rankings. Between a site at 92 and a site at 98, the second has a concrete SEO advantage on closely contested searches.

Mobile-first index. Google indexes and ranks the mobile version first, so the mobile Lighthouse score is the one that counts. Yet it is harder to optimise than desktop (a 4G connection, a less powerful processor, a smaller screen), which means reaching 95 on mobile takes more effort than 95 on desktop.

Competitive convergence. In 2026, many agencies have understood the importance of Lighthouse. Your direct competitors have raised their scores. If you stay at 75 while they are at 95, the gap shows in the rankings.

Anatomy of a poor score: the typical causes

A mobile Lighthouse score of 65-75 (typical of SMEs in French-speaking Switzerland in 2026) often combines the following 5 problems.

Problem 1: Images that are too heavy

Symptom: LCP > 4 seconds, low Performance score.

Cause: PNG/JPEG images without compression, without a modern format (WebP/AVIF), without lazy loading, without declared dimensions.

Solution: convert all images to WebP or AVIF (30-50% smaller files), serve them responsively with srcset, lazy load everything below the fold, declare width and height to avoid CLS.

Tools: Squoosh.app (manual compression), Imagify or ShortPixel (automatic for WordPress), the Astro/Next.js image service (automatic at build time).

Problem 2: Blocking JavaScript

Symptom: TBT (Total Blocking Time) > 600 ms, degraded INP.

Cause: third-party scripts loaded synchronously (Google Tag Manager, Hotjar, chat widgets), a main JavaScript bundle that is too big (> 200 KB), no code splitting.

Solution: switch all third-party scripts to async or defer. Code split the main bundle into logical chunks. Assess each third-party script: is it really useful?

Common mistake: loading Google Tag Manager on 100% of the site just to track 3 events. Prefer direct tracking without GTM where relevant.

Problem 3: Blocking web fonts

Symptom: FOIT (Flash Of Invisible Text) at the start of loading, low Performance score.

Cause: custom fonts loaded synchronously from Google Fonts or Adobe Fonts.

Solution: font-display: swap in @font-face, font subsetting (load only the characters you use).

Problem 4: Render-blocking CSS

Symptom: FCP (First Contentful Paint) > 1.8 s.

Cause: all of the site’s CSS loaded synchronously in the <head>.

Solution: extract the critical CSS (visible above the fold) and inline it in the HTML, load the rest in a non-blocking way.

Tools: Critters (Astro), criticalCSS (Next.js), penthouse.

Problem 5: WordPress plugins piling up

Symptom: Performance score < 60 even on simple sites.

Cause: 20-40 WordPress plugins loaded, each adding 10-50 KB of JS and server latency.

Solution: audit your plugins, deactivate the non-essential ones, replace stacks of plugins with native solutions. Or migrate to a modern stack (see our WordPress SEO article).

The 9 concrete actions to go from 70 to 95+

Action 1: Audit the current state

PageSpeed Insights → enter your URL → note the scores and the opportunities Lighthouse identifies. Test mobile and desktop, even though mobile is what counts.

Action 2: Image compression

The first high-impact action. Systematic WebP/AVIF conversion. On Astro/Next.js, it is automatic via <Image>. On WordPress, Imagify or ShortPixel do the job (CHF 5-10/month).

Action 3: Lazy loading

For all images below the fold: native loading="lazy". For embedded YouTube videos, use lite-youtube-embed rather than the standard embed.

Action 4: Declared image dimensions

Always set width and height on every image, either in HTML or with CSS aspect-ratio. This avoids CLS (layout shift during loading).

Action 5: JavaScript code splitting

Split the JS bundle into chunks per route. On Next.js, this is native. On Astro, JS is minimal by default (islands only). On WordPress with a heavy theme, it is more complex.

Action 6: Third-party scripts in async or defer

All non-critical <script> tags in async or defer. Assess whether each third-party script is necessary (chat widget, analytics, Google Tag Manager).

Action 7: Optimised fonts

font-display: swap on every @font-face. Self-host if possible. Subset custom fonts to reduce their size.

Action 8: Inlined critical CSS

The CSS for the first viewport inlined in the <head>. The rest loaded in a non-blocking way.

Action 9: Schema.org for the Lighthouse SEO score

The Lighthouse SEO score checks for basic schema.org. Implement Organization, BreadcrumbList and Article at a minimum.

Mistakes to avoid

Mistake 1: optimising desktop only. Many SMEs test only on desktop and conclude “my Lighthouse is at 95”, while mobile stays at 60, and Google ranks on mobile.

Mistake 2: measuring only once. Lighthouse varies with conditions (cache, network, server load). Run 3 tests and take the median.

Mistake 3: ignoring the new metrics. INP replaced FID in 2024. Many outdated articles do not mention INP. Make sure your audit uses the 2026 metrics.

Mistake 4: a score artificially inflated by removing things. Disabling all your features to inflate the Lighthouse score does not help if your site becomes useless. Keep the user experience in mind.

What Voren delivers

Every Voren site is delivered with a mobile Lighthouse score measured at go-live. It is our delivery standard:

  • Business website, One page and Essential: Lighthouse > 95 mobile + desktop
  • Business website, Premium: Lighthouse > 95
  • Business website, Bespoke: Lighthouse > 95
  • E-commerce: Lighthouse > 90 (e-commerce sites are structurally more complex to optimise)
  • Web app: Lighthouse > 85 depending on complexity

Default stack: Astro 5 (editorial sites) or Next.js 15 (interactive sites) + optimised images + async scripts + schema.org + inlined critical CSS.

Going further

Request a free Lighthouse audit: a complete diagnosis of your current site and an action plan within 7 days.

See our business website service: Lighthouse 90-95+ targeted and measured at go-live.

Read our Core Web Vitals article: understand the 3 metrics weighted in Lighthouse Performance.

Clovis, founder of Voren Clovis · founder

Let’s build what’s next.

Tell us where you want to go. We send back a mockup and a clear price, fixed before we start.

Call