What is a good Lighthouse score?

A Lighthouse score of 90–100 is good, 50–89 needs improvement and 0–49 is poor. The performance score is a weighted blend of lab metrics — Total Blocking Time, Largest Contentful Paint and Cumulative Layout Shift carry most of the weight — measured under simulated mobile conditions, so it varies between runs.

Last updated 2026-07-28

What counts as a good Lighthouse score?

90 or above is good, 50 to 89 needs improvement, and anything under 50 is poor. Those bands apply to each of the four Lighthouse categories separately — performance, accessibility, best practices and SEO — so a site can be green on three and red on the fourth.

The bands come from a log-normal distribution of real site performance rather than an arbitrary curve. Lighthouse anchors the scale against HTTP Archive data so that roughly the 25th percentile of sites lands on 50 and the top few per cent land on 90, which is why moving from 40 to 50 is far easier than moving from 90 to 95.

What goes into the Lighthouse performance score?

Five lab metrics, weighted. Under the weighting Google currently documents, Total Blocking Time contributes 30%, Largest Contentful Paint 25%, Cumulative Layout Shift 25%, First Contentful Paint 10% and Speed Index 10% — so the three heavyweight metrics decide four fifths of the number between them.

These weights have changed between major Lighthouse versions before, and audits are added and retired with Chrome releases. That is the everyday explanation for a score that drops several points overnight on a site nobody touched: check the Lighthouse version in the report footer before hunting for a regression.

Two of the three heavyweights are Core Web Vitals you can read about in detail — Largest Contentful Paint and Cumulative Layout Shift. Total Blocking Time is the lab stand-in for responsiveness: it sums the parts of long main-thread tasks that exceed 50 milliseconds, and it is the metric most sensitive to heavy third-party JavaScript.

Does 100 in accessibility mean the site is accessible?

No. A perfect accessibility score means no automated check failed, which is a much smaller claim than "this site is usable by everyone". Automated tooling can only detect a minority of real accessibility barriers — anything requiring judgement, such as whether alt text is meaningful or whether a keyboard journey makes sense, is invisible to it.

The same caveat applies to the SEO and best practices categories. They are pass/fail checklists rather than weighted measurements, so a 100 in SEO means the mechanical checks passed — a title exists, the page is crawlable, the viewport meta tag is present — not that the page deserves to rank.

Treat green in these three categories as a floor rather than a finish line. The value is in the specific failing audits, not the aggregate.

Why does my Lighthouse score change every time I run it?

Because it is a live measurement of one page load on your machine, and the conditions differ every time. Two runs a minute apart can differ by ten points or more on a page with third-party scripts, without anything on the site having changed.

• Network variance and CPU contention on the machine running the test.
• Third-party scripts — ads, chat widgets, analytics — responding at different speeds.
• A/B tests or personalisation serving different content.
• Chrome and Lighthouse version changes altering weights or adding audits.
• Browser extensions injecting scripts, which is why headless or clean-profile runs are more stable.

Does the Lighthouse score affect Google rankings?

No. Google uses Core Web Vitals field data from real Chrome users as a ranking signal, not the Lighthouse score, which no Google system ever sees. Lighthouse is a diagnostic that helps you improve the underlying metrics, and it is entirely possible to pass Core Web Vitals with a mediocre Lighthouse score.

That said, three of the five performance metrics overlap with what field data measures, so a page that scores badly in the lab usually has something real to fix. The score is a useful smoke alarm and a poor scoreboard. See how field assessment actually works for what Google does measure.

How much should you care about the number?

Chasing 100 is usually a poor use of time — the gap between 45 and 85 is where users feel a difference, and the gap between 92 and 98 is mostly for the screenshot. Spend the effort where a person would notice it.

A single score is also a weak unit of analysis. One URL tested once tells you almost nothing; the same metric across two hundred pages tells you which template is at fault. Before-and-after runs on the same machine, in the same conditions, are the only comparison worth trusting.

This is the reasoning behind Audra's single Audra Score, which blends the Lighthouse categories across every crawled page — performance 30%, SEO 30%, accessibility 25%, best practices 15% — into one figure for site health rather than one figure for one URL. Audra runs Lighthouse, axe-core, broken-link checks and its own HTML SEO and answer-engine checks locally on macOS or Windows against your own installed Chrome, crawling 100 pages by default with configurable include and exclude globs, and exports a self-contained HTML report or a print-ready PDF. It is $19 once with a free 7-day trial, no subscription. For a straight comparison with running Lighthouse yourself, see Lighthouse as a whole-site auditing tool.