Lightscore

What is a good Lighthouse score?

Lighthouse colours all four category gauges from one pair of thresholds. 0 to 49 is red. 50 to 89 is orange. 90 to 100 is green. Chrome's documentation covers the performance score and says sites "should strive to have a good score (90-100)".

I read five of the pages that rank for this question. None of them gives the next number. On mobile, the median page in the HTTP Archive crawl scores 49 for performance. Half the crawl is red.

What do real pages score?

HTTP Archive runs Lighthouse against millions of pages every month and publishes the median per category. From the July 2026 crawl:

Category Mobile Desktop
Performance 49 65
Best practices 77 77
Accessibility 86.5 86.5
SEO 92 92

The median page already clears 90 on SEO. It's 41 points short on mobile performance. Each category asks for a different amount of work to reach the same band.

Lighthouse 10 landed on 9 February 2023. It dropped Time to Interactive from the score and raised Cumulative Layout Shift from 15 to 25 percent. The crawl's median steps up between the February and March crawls, so numbers from either side of that break aren't comparable. Since March 2023 the mobile median rose from 38.5 to 49. It is still below 50.

Median performance score 0 20 40 60 80 100 2023-03 2023-09 2024-03 2024-09 2025-03 2025-09 2026-03 2026-07 Mobile Desktop Good threshold
Median Lighthouse performance score across the HTTP Archive crawl, every three months from the move to Lighthouse 10 scoring, plus the July 2026 crawl. Both curves flatten well short of the green band.

Only one of the four scores is a measurement

We keep three example reports warm, so the same three pages get measured over and over. Between 12 July and 14 August 2026 (UTC) that produced 132 single-run mobile results, all on Lighthouse 12.8.2. Frankfurt and Washington D.C. produced 121 of them, because those are the two regions the warm job books. The other 11 are visitor-chosen regions that landed in the same 24-hour cache. Performance shows the median and the full range.

Page Performance Accessibility Best practices SEO
lightscore.dev 100 (98–100) 100 100 100
wikipedia.org 92 (78–97) 96 100 100
vercel.com 32 (30–40) 95 93 92

The last three columns carry no range because they never moved. Each page returned the same accessibility score, the same best-practices score and the same SEO score on all 132 runs.

Chrome's documentation says why for accessibility. "Each accessibility audit is pass or fail," and the category score is a weighted average of those checks. Buttons either have accessible names or they don't. A slow network can't change the answer. Best practices and SEO behaved the same way in our log.

So three of the four scores tell you whether you did the work. Performance tells you how fast a machine was on the day. I read those as two different kinds of number.

The vercel.com row is the useful one. A company that sells web infrastructure puts three scores above 90 and one at 32, on its own homepage. A single number called "the Lighthouse score" hides that completely.

Is 90 realistic on mobile?

Lighthouse scores each metric on its own curve, and the value that earns a metric a 90 is a constant in the source. These are the constants in 12.8.2, the version our fleet pins:

Metric Mobile Desktop
First Contentful Paint 1,800 ms 934 ms
Largest Contentful Paint 2,500 ms 1,200 ms
Speed Index 3,387 ms 1,311 ms
Total Blocking Time 200 ms 150 ms
Cumulative Layout Shift 0.1 0.1

Desktop gets a tighter threshold on four of the five. The two paint metrics roughly halve. Speed Index drops to under 40 percent of the mobile value. Total Blocking Time only tightens by a quarter, and Cumulative Layout Shift doesn't move at all.

Desktop still scores higher. Lighthouse throttles the mobile run to about 1.6 Mbps and slows its CPU by a factor of four. The desktop run keeps its full-speed CPU and gets about 10 Mbps.

Total Blocking TimeHow long the main thread was busy and couldn’t respond to taps or clicks. carries 30 percent of the performance score, more weight than any other metric. The 2025 Web Almanac puts the median mobile value at 1,916 ms. That's nearly ten times the threshold for a 90, on the metric that counts most.

So what should you aim at?

  1. Aim at a metric, not the composite. The composite hides which of the five metrics moved. Start with Total Blocking Time and Largest Contentful Paint, which carry 55 percent of it between them.
  2. Fix the device and keep it fixed. The two profiles use different thresholds and different throttling, so a score without a device attached means nothing.
  3. Judge against a spread, not one run. One run tells you less than you think — see why your Lighthouse scores keep changing.
  4. Don't read the score as a Core Web Vitals verdict. Those are assessed on field data, at the 75th percentile of real visits.
what’s this

The small set of field metrics Google treats as a ranking input and reports in Search Console. We measure their lab equivalents here — real-user field data (CrUX) is a separate signal. TBT is the lab stand-in for INP, not a Core Web Vital itself.

Every free Lightscore report returns all four category scores per region, with the raw Lighthouse JSON and HTML behind them. On the API you pick the categories, and all four is the default. What none of it tells you is whether 92 is good enough for your page. That one's yours.

Common questions