Website Speed Test on Google's own engine
Enter a URL; Lighthouse runs on it and you receive four scores, the three key page-experience metrics and improvement opportunities sorted by their effect. Running mobile and desktop at once is possible.
- The LCP, CLS and TBT metrics
- Both device types
- Opportunities ordered by effect
Run the tool right here on the page
The same score Google assigns for speed, SEO, accessibility, and best practices
- What it costs $0.011
- Data source DataForSEO
The short answer
A site speed test means running Lighthouse, the very engine PageSpeed Insights is built on, to learn how long your page needs before it is usable. The figures that truly count are LCP, CLS and TBT, not the large coloured score sitting at the top.
The zero-to-one-hundred score is a weighted average of several metrics, and that is what makes it misleading. Two sites can reach the same score when one has an oversized-image problem and the other a heavy-JavaScript problem, and those two fixes have nothing in common. To act, read the three metrics below the score, not the score.
LCP marks when the largest element of the first screen showed up, and that element is nearly always an image or a large text block. CLS measures how far elements shifted after loading, the thing that makes you tap the wrong button; its usual cause is an image with no width and height. TBT reports how long the browser was tied up running JavaScript instead of answering your tap.
The order in which you address the three also matters, and it is usually done backwards. Start with CLS, the cheapest, which mostly comes down to setting dimensions on images and holding space for banners. Then LCP, which most often means shrinking and optimising the large image at the top. Save TBT for last, because it usually means removing or deferring scripts, and that is where something can break.
The output you receive
The exact meaning of each output column. Read this before your first run so you do not misread a value.
- LCP (Largest Contentful Paint)
- How long the largest content element of the page takes to paint; green below 2.5s, yellow through 4s, red above.
- CLS (Cumulative Layout Shift)
- How much the elements jump unexpectedly while the page loads; green below 0.1, yellow through 0.25, red above.
- TBT
- The field metric INP depends on real Google user data and is not obtainable in a laboratory run; in its place the report shows Total Blocking Time, the nearest lab-side equivalent for page responsiveness.
The data source and method
Lighthouse is a laboratory measurement, not a field one. It executes on a server with a simulated network profile, so its score sits a little apart from what actual Google users experience in the CrUX dataset, and it drifts a few points between one run and the next as a matter of course.
The situations where it helps
Right before and right after any technical change to the site, such as swapping the theme, installing a plugin, or compressing images, so the change's effect on speed is measurable.
Where it falls short
Lighthouse is a laboratory measurement, not a field one: it runs on a server with a simulated network, and its number is not necessarily what your real visitor gets on their own phone and connection. Two runs in a row against one URL also drift a few points on their own, so a three-or-four-point difference between runs means nothing and chasing it wastes time. What Google uses for ranking is field data from real users, seen in Search Console, not here. This tool exists to find the cause, not to hand out a grade.
A lesson from working with it
The trap almost every site walks into is lifting the score instead of making the site fast, and the two are not the same. Deferring every script until after user interaction raises the score while that same user now waits longer for buttons to react. Given a choice between a higher score and a page that becomes usable sooner, take the second.
What it costs
There is no monthly subscription. You raise the wallet balance, and the price of each run is written on the button before you press it; without a run there is no deduction.
Common questions about this tool
What score is good?
Instead of aiming at a score, look at Google's own thresholds on the three metrics. Those are the numbers Google has set good, needs-improvement and poor bands for, and they are colour-coded in this result. A page with all three metrics in the good band is healthy even with a middling overall score.
Why is my mobile score so much lower than desktop?
Because the mobile test deliberately models a weaker device and a slower network. That is normal, and every site shows the same gap. For a meaningful comparison, measure mobile against mobile, not against desktop.
How much does site speed affect rankings?
There is an effect, but not the one the marketing claims. In our experience, speeding up a page by itself usually shifts the ranking a little, while it improves bounce rate clearly. Fix speed for the visitor, not for Google; the SEO gain arrives on its own.