Three tools dominate free page speed testing: Google's PageSpeed Insights, GTmetrix and WebPageTest. They all produce numbers and recommendations, but they do not measure the same thing, so the same URL can get three different answers. This page explains what each tool is built for, why scores disagree, and how to use them together without chasing a single number.
Product details come from Google's PageSpeed Insights documentation, GTmetrix's features page and WebPageTest's documentation, read in October 2026. Features and plans change, so check each site for the current state. Where we mention SiteDownCheck, we are one more option, not a replacement for any of them.
PageSpeed Insights: field data plus a lab run
Google's documentation says PageSpeed Insights (PSI) reports two kinds of data. Field data comes from the Chrome User Experience Report (CrUX) and covers real Chrome users over the previous 28 days: First Contentful Paint, Interaction to Next Paint, Largest Contentful Paint, Cumulative Layout Shift and an experimental time to first byte. Lab data comes from Lighthouse, a simulated load that covers performance, accessibility, best practices and SEO. Mobile tests simulate a mid-tier phone on a mobile network, and desktop tests use an emulated desktop on a wired connection.
PSI's strength is that it is the closest free view of what Google itself sees. If the field data says your LCP is poor for real visitors, that is the number tied to the Core Web Vitals assessment. When a page has too little traffic, PSI falls back to origin-level data, and if the origin also has too little, there is no field data at all. Its limit is control: you cannot pick a location or throttle the connection.
GTmetrix: Lighthouse in a real browser with a UI for debugging
GTmetrix says each test loads the page in a real Chrome browser, with a Lighthouse-powered performance score combined with a structure score. Its features page lists 28 test locations, over 55 simulated devices, connection speeds from unthrottled to 3G or custom, a waterfall, filmstrip and video of the load, and scheduled monitoring with alerts. Free accounts exist; higher tiers add locations, monitored slots and retention. Check the current pricing page for the details.
GTmetrix is the friendliest of the three for a person who wants to see what happened and in what order. It is also useful for tracking a page over time.
WebPageTest: the deepest control
WebPageTest's documentation describes test machines around the world, a choice of browsers (Chrome, Firefox, Edge and others, varying by location), emulated mobile devices, a filmstrip that can compare runs, request and connection waterfalls and repeat views to measure a warm cache. The docs are now branded as Catchpoint's WebPageTest documentation, and they note that most Opportunities and Experiments require WebPageTest Pro while some are free. Verify the free versus paid split on their current site.
It rewards people who know what they are looking for. Comparing first view against repeat view, testing from a region close to your users, or running an experiment that blocks one third-party script are things it does well.
Side-by-side
| PageSpeed Insights | GTmetrix | WebPageTest | |
|---|---|---|---|
| Main data | CrUX field data plus Lighthouse lab run | Lighthouse-based lab test in real Chrome | Lab tests with detailed waterfalls and filmstrips |
| Real-user data | Yes, 28-day CrUX when available | INP from CrUX when available, per GTmetrix | Not its focus |
| Choose location | No | Yes (more on higher tiers) | Yes |
| Choose browser/device | Mobile or desktop | Many simulated devices | Several browsers and emulated devices |
| Waterfall | Limited audit view | Yes | Yes, with connection view |
| Repeat view / cache test | No | Check their docs | Yes |
| Scheduled monitoring | No | Yes | Check their current plans |
| Learning curve | Low | Low to medium | Medium to high |
Why the scores disagree
- Different machines and networks. Each tool throttles the CPU and network its own way. A Lighthouse score from one location will not equal one from another.
- Field versus lab. PSI field data reflects what real users experienced over weeks. Lab tests are one simulated visit right now.
- Variance. Even identical settings produce different results between runs, due to server load, third-party responses and A/B tests.
- Cache state. A first view and a repeat view are different experiences. Check which one a tool measured.
Pick one tool and one set of conditions for before-and-after comparisons, and treat the other tools as second opinions.
Where SiteDownCheck fits
Our Core Web Vitals tool runs Google's PageSpeed Insights engine for a URL, so it is a quick way to get a Lighthouse score and field data without leaving a diagnostics page. Our website speed test is different: it times DNS, connection, TLS and time to first byte for a single server response and does not render the page. It cannot replace either of the three tools above for front-end work. See the GTmetrix alternative and WebPageTest alternative pages for the details.
A practical order of use
- Start with PSI or our Core Web Vitals tool to see whether field data exists and which metric fails.
- If TTFB looks high, run the website speed test and fix the backend or caching first.
- Use GTmetrix when you want a readable waterfall and a record of changes over time.
- Use WebPageTest when you need a specific location, browser, repeat view or an experiment.
- Check weight and scripts with the page weight analyzer and third-party script checker.
- Confirm delivery with the HTTP/2 and HTTP/3 checker and CDN detector.
Honest limits
None of these tools can tell you how your site feels on your visitors' actual devices unless field data exists for your site. Small sites often have no CrUX data, in which case you only have lab numbers. Our tools add server timing and a PSI wrapper, but not a browser of our own.
Next steps
- Run Core Web Vitals on your most visited page.
- Time your server with the website speed test.
- Read the Core Web Vitals guide.
- Work through website slow diagnosis.
- Learn the difference in synthetic vs real user monitoring.
FAQ
Which of the three is the most accurate?
None is universally more accurate. PSI field data best reflects real users, while GTmetrix and WebPageTest give controllable lab conditions. Use the one that answers your question.
Why does my PSI score differ from GTmetrix?
Different test locations, throttling, devices and run-to-run variance. Compare results only from the same tool and settings.
Do I need field data to pass Core Web Vitals?
Google assesses real-user data when it exists. If your site has too little traffic for CrUX, only lab results are available, so improve those and recheck later.
Are these tools free?
All three have free options at the time of writing, with paid tiers adding features. Confirm limits on each site.
Try SiteDownCheck free
3 monitors, 5-minute checks, email and webhook alerts and a public status page. No credit card.
Start free monitoringOpinions here are our own; verify vendor details yourself. See our editorial policy and disclaimer.