What a 100/100 Lighthouse score actually means
This site scores 100 on all four categories. Here is what that does and does not tell you, using our own numbers as the worked example.
Agencies like to screenshot a perfect Lighthouse score. It is a satisfying image and it is frequently oversold, so here is ours along with an honest account of what it is worth.
Our numbers
Measured with Lighthouse CI on 3 September 2026, three runs per page, median reported, desktop preset:
Every page: 100 performance, 100 accessibility, 100 best practices, 100 SEO.
The homepage specifically:
- Largest Contentful Paint: 0.57s — the main content appears in just over half a second. The threshold for "good" is 2.5s.
- Cumulative Layout Shift: 0.000 — nothing moves while the page loads.
- Total Blocking Time: 0ms — the page is never busy enough to ignore a tap.
- JavaScript: about 143 KB on the homepage.
Two caveats immediately, because they matter.
This is the desktop preset. Mobile scores are always lower — the test simulates a slower processor and connection. Anyone showing you a perfect score without saying which preset it was is showing you the flattering one.
This is a lab test, not real users. Lighthouse runs on one machine under simulated conditions. It is a useful proxy, not a measurement of what your customers experience.
What the score genuinely tells you
That the basics were not skipped. No enormous unoptimised images, no render-blocking scripts, headings in a sensible order, images with alt text, sufficient colour contrast, a description on every page. Boring things, routinely missing.
That someone measured. Most sites have never been tested. A perfect score means somebody looked, which correlates with a lot of other care.
That nothing is grossly wrong. Lighthouse catches broken basics reliably.
What it does not tell you
This is the more important half.
It does not mean the site is accessible. A 100 accessibility score means it passed the automated checks, and automated tools catch only a portion of real accessibility problems. They cannot tell whether your alt text is meaningful, whether the tab order makes sense, or whether the page is usable with a screen reader. Those need a person. A 100 is a floor, not a certificate.
It does not mean the site is fast for your customers. Lab conditions on a good machine. Real people are on old phones on poor connections. Field data — from Search Console or the Chrome UX Report — is what actually counts, and it is always less flattering.
It does not mean the site works. A blank page with a heading scores 100. So does a beautifully fast website nobody understands or converts on. Lighthouse has no opinion about whether your copy makes sense, whether people can find your phone number, or whether the thing sells.
It is not a ranking. Page speed is one input among many. A 100 will not lift you above a slower competitor with better content and more links.
And it is a moment in time. Scores drift as content is added. One good screenshot from a launch two years ago says nothing about today.
Why we bother
Not for the screenshot.
The score is enforced in our build system: the same checks run on every change before it can ship, and if a change makes the site slower, breaks contrast, or drops a page description, the build fails and the change does not go out.
That is the useful part. Not the number — the fact that it cannot quietly get worse. Most sites are fast on the day they launch and slow eighteen months later, one addition at a time, and nobody notices because nobody is checking.
What you should ask for instead
If you are hiring someone and want to judge this properly, do not ask for a Lighthouse score. Ask:
- "What does your own site score on mobile?" Mobile is the honest number, and their own site is the sample they had complete control over.
- "Is speed checked automatically when you make changes, or only at launch?" This separates people who care from people who tidied up once.
- "What field data do you have?" Real user measurements beat lab tests.
- "Can you show me a client site you built two years ago?" Still fast, or degraded?
You can check the first one yourself before the call. Run PageSpeed Insights on any agency's own website — it takes thirty seconds, and it is remarkable how many people selling websites have a slow one.
Ours is the site you are reading. Test it.