How to Actually Get a 100 Lighthouse Score in 2026 (and Keep It)

To hit 100 in Lighthouse, fix the metrics it measures: serve correctly sized AVIF or WebP images, cut render-blocking JS, set explicit dimensions to stop layout shift, and pass every Accessibility, Best Practices, and SEO audit. Then keep it by testing on real devices, since lab scores don't reflect real-world speed.
Part of the complete web development guide.
On this page
- Fix the things Lighthouse actually measures
- 1. Serve properly sized, modern-format images
- 2. Cut render-blocking CSS and JS
- 3. Reduce Total Blocking Time
- 4. Get your fonts under control
- 5. Stop shipping unused code
- 6. Fix the accessibility and SEO audits too
- What to do when you’re stuck at 95
- FAQ
- Does a 100 Lighthouse score help my SEO?
- Why does my score change every time I run Lighthouse?
- Is it worth paying someone to optimize my site?
Lighthouse scores are useful, but chasing a perfect 100 for its own sake is a trap. Google’s scoring model has changed enough over the past few years that some of the old tricks people used to hit 100 now actively hurt real performance. What follows is what actually moves the needle in 2026, based on what I see auditing sites around Long Island and beyond.
One thing before the list: Lighthouse is a lab test. It runs on a simulated throttled connection and a specific device profile. A 100 in the lab does not guarantee a fast site for a real visitor on a real phone in a real parking lot with bad LTE. Use it as a diagnostic, not a trophy.
Fix the things Lighthouse actually measures
Lighthouse’s Performance score in 2026 weights these metrics, roughly in order of impact: Largest Contentful Paint (LCP), Total Blocking Time (TBT), Cumulative Layout Shift (CLS), First Contentful Paint (FCP), and Speed Index. Accessibility, Best Practices, and SEO are separate scores with their own audits. If you want a 100 across the board, you have to address all four categories, not just Performance.
1. Serve properly sized, modern-format images
This is still the single biggest win on most sites I audit. A hero image exported at 2400px wide and served to a 400px phone is the most common cause of a bad LCP. Use responsive images with explicit dimensions and modern formats.
<picture>
<source
srcset="/hero-800.avif 800w, /hero-1600.avif 1600w"
sizes="(max-width: 800px) 100vw, 800px"
type="image/avif">
<source
srcset="https://i0.wp.com/www.kennykey.com/hero-800.webp 800w, https://i0.wp.com/www.kennykey.com/hero-1600.webp 1600w"
sizes="(max-width: 800px) 100vw, 800px"
type="image/webp">
<img
src="https://i0.wp.com/www.kennykey.com/hero-800.jpg"
width="800"
height="450"
loading="eager"
fetchpriority="high"
alt="Storefront on Main Street">
</picture>
Note the width and height attributes. Those reserve space and prevent layout shift, which is a big part of your CLS score. Also note fetchpriority="high" on the LCP image. That tells the browser to load it before other resources, and it can shave hundreds of milliseconds off LCP.
Lazy-load everything below the fold, but never lazy-load your LCP image. That’s a common mistake that makes LCP worse, not better.
2. Cut render-blocking CSS and JS
Every stylesheet and synchronous script in your <head> delays first paint. If you’re running a WordPress site with a dozen plugins each enqueuing their own CSS, this is probably your problem.
<!-- Blocking: delays render -->
<link rel="stylesheet" href="/styles.css">
<!-- Better: inline critical CSS, defer the rest -->
<style>/* above-the-fold styles here */</style>
<link rel="stylesheet" href="/styles.css" media="print" onload="this.media='all'">
<noscript><link rel="stylesheet" href="/styles.css"></noscript>
On the JS side, use ES modules with defer or type="module" so parsing doesn’t block the main thread. If you have a large third-party script (chat widgets, analytics, tag managers), load it after the page is interactive.
3. Reduce Total Blocking Time
TBT is the metric people ignore and then wonder why they’re stuck at 78. It measures how long the main thread is blocked by long tasks during page load. Any task over 50ms counts against you.
Common culprits: large framework bundles, synchronous third-party scripts, and hydration on content that doesn’t need to be interactive. If you’re on a JavaScript framework, look into partial hydration or islands architecture. Shipping a full React app to render a blog post is a self-inflicted wound.
You can find long tasks in the Chrome DevTools Performance panel. Look for the red triangles. Each one is a chance to split work or defer it.
4. Get your fonts under control
Custom fonts cause two problems: they block text rendering, and they cause layout shift when the fallback swaps to the real font. Both hurt your score.
@font-face {
font-family: 'Inter';
src: url('/fonts/inter.woff2') format('woff2');
font-display: swap;
unicode-range: U+0000-00FF, U+0131, U+0152-0153;
}
Three things matter here. Use woff2 only (it’s universally supported now). Use font-display: swap so text renders immediately with a fallback. And subset with unicode-range so you’re not downloading glyphs for languages you don’t serve.
Even better: preload your primary font file and match your fallback font’s metrics so the swap doesn’t shift anything.
5. Stop shipping unused code
This is where a lot of “I hired someone and they used a page builder” sites fall apart. If your CSS file is 400KB and 90% of it is unused, that’s dead weight on every page load.
Run a coverage report in Chrome DevTools (Command Palette, then “Coverage”). It shows exactly what percentage of your CSS and JS is actually used on the current page. Anything under about 40% used is worth investigating. Tools like PurgeCSS can strip unused selectors at build time, though you need to configure it carefully so you don’t purge classes that are added dynamically.
6. Fix the accessibility and SEO audits too
If you want 100 across all four categories, don’t forget the non-performance audits. They’re usually easier wins:
- Every image needs meaningful alt text (or empty alt if decorative).
- Color contrast needs to hit WCAG AA. Lighthouse flags this directly.
- Every page needs a unique, descriptive
<title>and meta description. - Links need discernible text. “Click here” fails the audit.
- Your document needs a valid
langattribute and a logical heading order. - Tap targets need to be at least 48×48 CSS pixels with adequate spacing.
These are boring fixes, but they’re the difference between 96 and 100, and they genuinely help real users. For a business site, accessibility isn’t optional anymore. It’s also just good practice.
What to do when you’re stuck at 95
The last few points are almost always one of three things: a third-party script you can’t easily remove, a font loading issue, or a single large image. Open the Lighthouse report, expand the opportunities and diagnostics sections, and look at the estimated savings in milliseconds. Fix the biggest one first, then re-run.
One warning: if you’re optimizing a site that’s already live and generating revenue, test changes on a staging environment first. I’ve seen people break checkout flows chasing a score. If you’re not comfortable making these changes yourself, working with someone who does custom web development on a performance-first basis is cheaper than the lost conversions from a broken site.
FAQ
Does a 100 Lighthouse score help my SEO?
Indirectly, yes. Google uses Core Web Vitals (LCP, INP, and CLS) as a ranking signal, and those overlap heavily with what Lighthouse measures. But a 100 doesn’t guarantee rankings. Content relevance and backlinks still matter more. Treat performance as a tiebreaker, not a strategy.
Why does my score change every time I run Lighthouse?
Because it’s a lab test with simulated throttling, and your machine’s CPU load, browser extensions, and network conditions all affect the result. Run it three times in an incognito window and average the scores. For a more stable picture, use the Chrome UX Report (CrUX) data in PageSpeed Insights, which reflects real user measurements over 28 days.
Is it worth paying someone to optimize my site?
If your site is slow enough that visitors bounce, yes. The math is simple: a one-second delay in load time measurably reduces conversions on most e-commerce and lead-gen sites. If you’re a small business with a five-page site on a decent host, you can probably handle the basics yourself with this guide. If you’re running a larger site with a build pipeline, it’s usually worth the money.


