How to use the Website Speed Test
- Enter the full URL of the page you want to test, for example
https://example.com/pricing/. - Choose Mobile or Desktop. Mobile is the stricter test and a good default.
- Start the test. It usually takes 15 to 60 seconds, because the page is loaded in a real browser and measured from start to finish.
- Check the screenshot to confirm the page loaded as expected, then read the performance score and the five lab metrics.
- Work through the opportunities list, starting with the largest estimated savings.
Understanding your results
Performance score
A score from 0 to 100, calculated by Lighthouse from five lab metrics. Lighthouse's bands are 90 to 100: good, 50 to 89: needs improvement and 0 to 49: poor. In the current Lighthouse scoring model, Total Blocking Time carries 30% of the weight, Largest Contentful Paint and Cumulative Layout Shift 25% each, and First Contentful Paint and Speed Index 10% each.
Lab metrics
- First Contentful Paint (FCP): when the first text or image appears on screen.
- Largest Contentful Paint (LCP): when the largest visible element, usually a hero image or main heading, finishes rendering.
- Total Blocking Time (TBT): how long the main thread was blocked by long tasks during loading. TBT is the lab stand-in for Interaction to Next Paint, which can only be measured with real user interactions.
- Cumulative Layout Shift (CLS): how much visible content moved unexpectedly while the page loaded.
- Speed Index: how quickly the visible part of the page filled in overall.
Screenshot
The page as our browser saw it after loading. If it shows an error page, a login screen or a bot challenge, the scores describe that screen, not your real page.
Opportunities
Specific fixes Lighthouse found, such as unused JavaScript, oversized images or render-blocking resources, each with an estimated saving. Treat the savings as estimates for prioritizing, not guarantees.
Core Web Vitals thresholds
Google's Core Web Vitals are measured on real visits and judged at the 75th percentile, meaning at least three quarters of page loads should meet the target:
| Metric | Good |
|---|---|
| Largest Contentful Paint (LCP) | 2.5 s or less |
| Interaction to Next Paint (INP) | 200 ms or less |
| Cumulative Layout Shift (CLS) | 0.1 or less |
Our lab LCP and CLS use the same definitions, and TBT stands in for INP, so this test helps you find and fix problems. It cannot tell you whether your real visitors pass, because it contains no real-user data.
How this tool works
We run Google's open-source Lighthouse ourselves, on our own server, in a real headless Chrome browser. The mobile profile emulates a mid-range phone on a throttled, slow 4G connection. The desktop profile uses a desktop-sized screen with a faster connection and no device slowdown. Lighthouse loads the page once and records everything above.
This is lab data only. A lab test is one controlled page load on one simulated device, which makes it good for debugging and for comparing results before and after a change. Field data is different: it records what real visitors experienced on their own devices and connections over time. We do not collect or show field data. To see real-user Core Web Vitals for your site, use Google's PageSpeed Insights, which shows Chrome UX Report data when enough visitors are available, or the Core Web Vitals report in Search Console for pages on your own verified site.
Scores vary between runs even when nothing on the page has changed. Server response times, network conditions, third-party scripts and ads all shift from one load to the next. Test location matters as well: the network timing you see reflects the distance between our server and yours, so a site hosted far from our server may load more slowly here than it does for visitors close to it.
Results are cached for 24 hours, so testing the same URL and profile again within that period shows the stored result.
Tips
- Run the mobile test first; it exposes heavy JavaScript and large images most clearly.
- Find your LCP element and make sure it loads early: a properly sized, compressed image that is not lazy-loaded.
- Cut and defer JavaScript to lower Total Blocking Time, which usually improves real-world INP too.
- Set width and height on images, and reserve space for ads and embeds, to reduce layout shift.
- Remove unnecessary redirects in front of your pages; each one adds a round trip. The Redirect Checker shows them.
- Find out which platform a slow site runs on with the CMS Detector, then look for platform-specific fixes.
- Our guide on how to improve website speed walks through the most common fixes in order.
Frequently asked questions
Why does my score change every time I run the test?
Each run is a fresh page load, and server response times, network conditions and third-party content change between loads. Small differences are normal. Look at the typical result across a few runs, keeping in mind that repeat tests within 24 hours return the cached result.
Does this test show real-user Core Web Vitals?
No. It is a lab test only. For field data from real visitors, check Google PageSpeed Insights or the Core Web Vitals report in Google Search Console.
Why is my mobile score lower than my desktop score?
The mobile profile emulates a mid-range phone on a throttled, slow 4G connection, which makes heavy scripts and large files far more costly. Most sites score lower on mobile.
Why is there no INP in the results?
Interaction to Next Paint needs real user interactions, which a lab test does not have. Total Blocking Time is the lab measure that tracks it most closely, so reducing TBT is the best lab-based way to work on INP.
Why does my site look slower here than in other tests?
Lab results depend on the test location, the device profile and the connection throttling. Our server may be farther from your hosting than another tool's test machine, which adds network time. Compare results within one tool rather than across tools.
What is a good performance score?
Lighthouse treats 90 to 100 as good. For your visitors and for search, passing the three Core Web Vitals in real-user data matters more than any single lab score.