CSS

How to Actually Hit a Perfect 100 Lighthouse Score (2026 Guide)

By Ken Key Aug 13, 2026 4 min read Updated Oct 16, 2026

How to Actually Hit a Perfect 100 Lighthouse Score (2026 Guide)
The short answer

A perfect 100 means passing every audit in all four Lighthouse categories, not just scoring well. In practice you get there by cutting requests, serving a properly sized hero image, keeping Total Blocking Time under 200ms, and eliminating layout shift. Accessibility, Best Practices, and SEO are mostly checklists.

Part of the complete web design guide.

On this page

A 100 across all four Lighthouse categories is doable, but it’s less about tricks and more about removing things. Over the years I’ve audited a lot of sites, from small business pages in Commack to larger apps, and the sites that score perfect are almost always the boring ones: few requests, no layout shifts, and JavaScript that earns its place. This is the process I actually use.

First, understand what Lighthouse is measuring in 2026

Lighthouse runs a synthetic test in a throttled Chrome instance. It doesn’t measure your real users. It measures a simulated one on a slow 4G connection with a mid-tier CPU. That distinction matters, because you can score 100 and still have a site that feels slow for someone on a real phone, and vice versa.

The four categories are Performance, Accessibility, Best Practices, and SEO. Performance is the hard one. The other three are mostly a checklist you can finish in an afternoon.

The performance score is three numbers

Lighthouse doesn’t score “speed” as one thing. It weights a handful of metrics, but three dominate:

  • Largest Contentful Paint (LCP) is the biggest weight. Target under 2.5 seconds.
  • Total Blocking Time (TBT) is your main-thread cost. Target under 200ms.
  • Cumulative Layout Shift (CLS) is visual stability. Target under 0.1.

If you fix those three, the score follows. Chasing the long tail of minor audits before these are green is a waste of time.

Fixing LCP: the image is usually the problem

On most sites the LCP element is a hero image or a heading. If it’s an image, the fix is almost always the same: serve it as AVIF or WebP, size it correctly, and don’t lazy-load it.

<img
 src="/hero-1200.avif"
 srcset="/hero-600.avif 600w, /hero-1200.avif 1200w"
 sizes="(max-width: 700px) 100vw, 1200px"
 width="1200"
 height="630"
 alt="Team working at a desk"
 fetchpriority="high"
 decoding="async">

Three things there do the heavy lifting. fetchpriority="high" tells the browser to grab it early instead of waiting. The explicit width and height reserve space so nothing jumps. And no loading="lazy" on the hero, because lazy-loading your LCP image is one of the most common self-inflicted wounds I see.

If the LCP element is text, the problem is usually the font. A web font that blocks rendering costs you. Use font-display: swap, self-host the font, and preload the one weight you actually use:

<link rel="preload" href="/fonts/inter.woff2" as="font" type="font/woff2" crossorigin>

Fixing TBT: ship less JavaScript

TBT is where modern sites fall apart. Every framework bundle, analytics tag, chat widget, and A/B testing script competes for the main thread. You can’t optimize your way out of a bloated bundle. You have to remove things.

Practical steps that work:

  • Open the Coverage panel in DevTools and find the JS you ship but never execute. Delete it.
  • Load third-party scripts after the page is interactive. A chat widget does not need to run before your content does.
  • If you’re using a framework, check whether you’re shipping hydration for components that never change. Static content doesn’t need to hydrate.
  • Audit your analytics. One good, lightweight script beats three overlapping ones.

Every script you cut is TBT you get back for free.

Fixing CLS: reserve space for everything

Layout shift comes from content that appears without the browser knowing its size ahead of time. The usual suspects are images without dimensions, ads, embeds, and fonts that swap late.

For anything you can’t size exactly (ads, embedded widgets), wrap it in a container with a fixed aspect ratio:

.ad-slot {
 aspect-ratio: 300 / 250;
 width: 100%;
 contain: layout;
}

The aspect-ratio reserves the box so nothing below it moves when the ad loads. The contain: layout tells the browser that changes inside this box won’t affect anything outside it. That one property solves a surprising number of shift problems.

Accessibility, Best Practices, and SEO are checklists

These three are far more mechanical. You can clear them without much guesswork.

Accessibility: real alt text, color contrast above 4.5:1 for body text, form inputs with labels, a logical heading order, and a page that works with the keyboard alone. Lighthouse flags most of this directly.

Best Practices: use HTTPS everywhere, keep your console clean, avoid deprecated APIs, and set proper security headers. A stale library throwing errors will cost you here.

SEO: a title and meta description on every page, a valid lang attribute, crawlable links with real text, and a viewport meta tag so mobile renders correctly.

A note on chasing the number

A 100 is a nice target, but it’s a synthetic score. I’ve built sites for clients across Long Island web design projects where the real win was cutting load time in half, not the badge. Use Lighthouse to find the biggest problems, fix them, then check your real-user metrics in the field. If your field data looks good, the score is doing its job.

The honest version: 100 is achievable on a focused marketing site or a well-built blog. On a large web app with lots of interactive features, getting all four to 100 is often not worth the tradeoffs in functionality. Aim for a fast site, and let the score reflect that.

FAQ

Is a perfect Lighthouse score actually worth chasing?

Partly. It’s a good forcing function for cutting bloat, but the number itself isn’t the goal. Real-user metrics and how the site feels matter more than the badge.

Why does my score change between runs?

Lighthouse is a synthetic test with simulated throttling, so there’s always some variance. Run it a few times and look at the median, not one result. Test in incognito with extensions off.

Can I get 100 without removing JavaScript?

Sometimes, if the JS is small and runs after first paint. But on most sites the fastest path to a high TBT score is shipping less code, not optimizing what’s there.

Have a project like this?

I do Long Island web design for Long Island businesses. If you need this done right, let’s talk.