What Is a Good Lighthouse Score?

Ninety is good. A hundred is a hobby. And neither is the number Google ranks you on.

What Is a Good Lighthouse Score? — Troiana insight cover

In short

A Lighthouse performance score of 90 or above is good; Google colours it green and it corresponds to a page that loads quickly on a mid-range phone. Scores of 50–89 are orange and mean real users are waiting; below 50 is red. The other three categories, accessibility, best practices and SEO, should be at or near 100 because they are checklists rather than measurements. Lighthouse is a lab test on a simulated device; the numbers Google uses for ranking are Core Web Vitals from real visitors, which Lighthouse only approximates.

The four scores

Lighthouse reports four numbers out of 100, and they are not the same kind of number.

Performance is a weighted blend of timing measurements taken while loading the page on a simulated mid-range phone over a throttled connection: how quickly the main content appears, how long the main thread is blocked, how much the layout shifts. It is a measurement, it varies between runs, and it is the score people mean when they ask the question.

Accessibility, best practices and SEO are checklists. They test whether images have alt text, whether the page has a title and a viewport tag, whether links have discernible names, whether the page uses HTTPS. A site can and should score at or near 100 on all three, because each point lost is a specific, fixable defect the report names. A 100 here does not mean the site is accessible or well-optimised; it means the automated checks passed, which catches perhaps a third of real accessibility problems. It is a floor, not a certificate.

What the performance bands mean

Google's own bands: 90–100 green, 50–89 orange, 0–49 red. Green corresponds roughly to a page whose Largest Contentful Paint is around 2.5 seconds or better on the simulated phone, with little blocking time and minimal layout shift. That is a page a visitor on an ordinary device experiences as fast.

Orange is the band most real websites live in, and it covers a wide range. A 55 and an 85 feel very different. Within it, the individual metrics matter more than the composite: a page at 70 because of one large image is a quick fix; a page at 70 because of a megabyte of JavaScript is a project.

Red means visitors are waiting several seconds on phones, and a meaningful share of them leave before the page is usable.

Why 90, not 100

The score is non-linear. Going from 60 to 90 usually means fixing a few large problems: an unsized hero image, a render-blocking script, a font that hides text. Going from 90 to 100 means eliminating every third-party script, inlining critical CSS, and often removing things the business wants on the page, such as analytics, chat, or consent management. The last ten points cost more than the previous thirty and deliver almost no perceptible difference to a visitor.

A 100 is achievable, and our own static-site approach tends to produce it, but it is a consequence of building simply rather than a goal worth chasing on a site with legitimate third-party needs. Green and stable is the sensible target.

The number Google actually uses

Lighthouse is a lab test: one run, one simulated device, one simulated network, from wherever the test was launched. Google's ranking signal is Core Web Vitals measured from real Chrome users over the previous 28 days, reported in Search Console and in the top section of PageSpeed Insights as "field data".

The two disagree constantly. A page can score 95 in the lab and fail Core Web Vitals in the field because real visitors are on slower phones than the simulation, or in a region far from the server, or on a page variant the test never saw. It can also score 60 in the lab and pass in the field because most visitors are on fast connections with the page cached.

So: use Lighthouse to find what to fix, because it explains itself. Use field data to know whether it mattered, because that is what ranks. The real-user measurement is the one to report to whoever asked.

Why the score changes every run

Because it is a measurement of a simulation running on whatever machine launched it. A busy laptop, a browser extension, a different Chrome version, a variable third-party response, or an A/B test variant all move the number by several points. Treat a single run as noise. Run three to five and take the median, use the same conditions each time, and look at the underlying metrics rather than the composite when comparing before and after.

What a good score does not tell you

A green performance score says nothing about whether the page converts, whether the content is worth reading, or whether the site is accessible in practice. It is also gameable: a page that loads a blank shell quickly and then fetches everything scores well and feels slow. The score is a proxy for one aspect of quality, and a useful one, and it should be treated as exactly that.

When to stop

When the performance score is green across the main page templates on the median of several runs, the other three categories are at or near 100, and the field data in Search Console shows the site passing Core Web Vitals. At that point further work on the score is cosmetic, and the time is better spent on what the pages say.

If you are stuck in orange and cannot see which item on the report is the one that matters, book a call; reading a Lighthouse report is quicker with someone who has read a few hundred.

Common questions

Is a Lighthouse score of 80 good?

It is in the orange band, which Google labels as needing improvement, but it is usually one or two fixes away from green. Check the individual metrics: if Largest Contentful Paint is the problem, an image or font is typically responsible; if Total Blocking Time is, it is JavaScript. An 80 with passing field data in Search Console is fine in practice.

Does a Lighthouse score affect SEO?

Not directly. Google ranks on Core Web Vitals measured from real visitors, not on Lighthouse's lab score. A better Lighthouse score usually improves the field numbers, so the two move together, but a site is never penalised for a lab score and never rewarded for one. Fix what Lighthouse reports; judge success by Search Console.

Why is my Lighthouse score different on mobile and desktop?

Mobile mode simulates a mid-range phone with a throttled CPU and a slow 4G connection; desktop mode simulates a fast machine on broadband. The same page takes several times longer to process on the simulated phone, so mobile scores are lower, sometimes dramatically. Google indexes mobile-first, so the mobile score is the one to work on.

Is a 100 Lighthouse score worth it?

Rarely as a goal. The last ten points typically require removing third-party scripts and inlining assets, which costs engineering time and often features the business wants, for a difference visitors cannot perceive. Aim for a stable 90+ on the median of several runs, and let 100 be a side effect of building simply rather than a target.

What is a good Lighthouse accessibility score?

100, or as close as the report allows. The accessibility category is a checklist of automated tests, each of which names a specific defect, so every lost point is fixable. A perfect score does not mean the site is fully accessible, because automated tests catch only a fraction of real barriers, but anything less means known problems are unfixed.

Have something worth building right?