Website Speed Test Tools Compared: PSI, Lighthouse, GTmetrix

What does a website speed test actually measure?

Every speed tool gives you one or both of two kinds of data. According to Google's PageSpeed Insights documentation, lab data is collected in a controlled environment and is useful for debugging, but it may miss real-world bottlenecks. Field data captures what real visitors experienced, with a smaller set of metrics.

The field metrics that matter most are the Core Web Vitals: Largest Contentful Paint (LCP) for loading, Interaction to Next Paint (INP) for responsiveness and Cumulative Layout Shift (CLS) for visual stability. Google's thresholds for a "Good" result are:

• LCP: 2,500 ms or less
• INP: 200 ms or less
• CLS: 0.1 or less

These are judged at the 75th percentile of visits, so a page passes only when most visitors get a good experience, not just the ones on fast phones.

Website speed test tools side by side

For GTmetrix plan options, check the official GTmetrix pricing page rather than any figure quoted in a blog.

PageSpeed Insights: start here for real-user data

Enter a URL and PageSpeed Insights shows two sections. The top one is field data from the Chrome User Experience Report: real users' FCP, INP, LCP and CLS over the previous 28 days, plus Time to First Byte as an experimental metric. If a page is new or has too few visits, PSI falls back to data for the whole origin, and sometimes it has no field data at all.

The lower section is a Lighthouse lab run that scores Performance, Accessibility, Best Practices and SEO (search basics). On mobile it simulates a mid-tier device (a Moto G4) on a mobile network. A score of 90 or above is good, 50 to 89 needs improvement and below 50 is poor.

Two things trip owners up. First, Google itself says a green lab score does not guarantee good real-user experiences. Second, the performance score moves from run to run even when nothing changed, because of network, hardware and resource contention. Read the field section first; use the lab section to find out why.

Lighthouse: the lab test you run while fixing

Lighthouse is an open-source tool you can run in Chrome DevTools, from the command line, as a Node module or through PageSpeed Insights. The DevTools version is the one we use most, because it can audit pages that need a login and sites running locally, such as a checkout, an account area or a staging build. A report takes about 30 to 60 seconds.

For repeat checks, the command-line version needs Google Chrome installed:

npm install -g lighthouse
lighthouse https://example.com/

Teams that deploy often can add Lighthouse CI to catch regressions before they go live. Lighthouse also has "stack packs" that detect the platform a page is built on and show advice specific to it.

One caution: a run on your laptop reflects your laptop. Local network, hardware and other open tabs all change the numbers, so compare runs on the same machine under the same conditions.

GTmetrix: waterfalls, regions and monitoring

GTmetrix gives each page a letter grade from A to F. The grade is a weighted average of two scores:

• Performance Score (60%): essentially your Lighthouse performance score, captured with GTmetrix's own browser, hardware and analysis options such as connection speed. Its weights are LCP 25%, Total Blocking Time 30%, CLS 25%, First Contentful Paint 10% and Speed Index 10%.
• Structure Score (40%): GTmetrix's own assessment of Lighthouse audits, plus custom checks such as CDN use and avoiding CSS @import.

Where GTmetrix earns its place is the extra tabs. The waterfall chart breaks the page load down request by request, which is the fastest way we know to spot one oversized image or a slow third-party script. You can also pick the test region, record a video of the load, view history graphs and set alerts on monitored pages.

Why do the three tools give different scores?

GTmetrix explains that its performance score and PSI's are not directly comparable, though they should be similar. The usual reasons are:

• Location: PSI runs in a Google data centre; GTmetrix lets you choose a region.
• Hardware and network: each tool uses different CPU, memory and connection settings.
• Lab versus field: a single simulated load on one device will not match 28 days of real visitors on many devices.
• Normal variance: scores shift between runs even with no code change.

So don't chase one number across tools. Pick one tool and one set of settings for before-and-after comparisons, run it a few times, and watch the individual metrics (LCP, CLS, Total Blocking Time) rather than the headline score.

The order we use on a slow client site

1. PageSpeed Insights, mobile first, on the homepage and one key product or service page. Note whether field data exists and whether it passes Core Web Vitals.
2. GTmetrix waterfall, from the region closest to most customers, to find the heaviest or slowest requests.
3. Lighthouse in DevTools while we make changes, including on logged-in pages and staging.
4. Re-test after deploy. Lab results change straight away. PSI's field data is updated daily but covers a trailing 28-day window, so real-user numbers improve gradually.

On Shopify, the fixes usually start with images, apps and theme code; our guide to Shopify speed optimisation without a rebuild walks through them. If the platform itself is the bottleneck, our comparison of Next.js vs WordPress for a business website explains when a different stack makes sense. And if you'd rather someone keeps an eye on this every month, that is part of our website maintenance service.

Frequently asked questions

Which website speed test tool is most accurate?

None of them is "the" accurate one, because they measure different things. Field data in PageSpeed Insights (or the GTmetrix CrUX tab) shows what real visitors experienced over 28 days. Lab tools such as Lighthouse and GTmetrix's tests are for diagnosing and checking fixes.

Why does my PageSpeed Insights score change every time I test?

Google lists local network availability, client hardware and resource contention as common causes of variability. Run the test a few times and look at the trend, not one result.

Do I need a score of 100?

No. PSI treats 90 or above as good, and Google notes that good lab data does not guarantee good real-user experiences. Passing Core Web Vitals for real visitors matters more than the last few lab points.

Can I test a page that needs a login?

Yes. Google's Lighthouse documentation says the Chrome DevTools workflow can audit pages that require authentication, as well as sites running locally. That makes it the right tool for checkouts, account areas and staging builds.

Is GTmetrix better than PageSpeed Insights?

They do different jobs. PSI is the quickest way to see Google's real-user data; GTmetrix is stronger for waterfalls, regional tests and tracking a page over time. We use both.