# Lighthouse vs PageSpeed Insights: lab and field

Lighthouse is the lab test you run. PageSpeed Insights runs Lighthouse and adds field data from real Chrome users. Use both: Lighthouse to fix causes, PageSpeed Insights to judge impact and report Core Web Vitals.

Updated 2026-09-13 · Source: https://porteur.ai/compare/lighthouse-vs-pagespeed-insights

## The short answer

Run Lighthouse when you need a controlled, repeatable test of one page. Use PageSpeed Insights when you need real-user data and a Core Web Vitals verdict.

Scores differ because PageSpeed Insights shows two things: a Lighthouse lab score and Chrome UX Report field data. Lighthouse alone shows only lab results. Lab results are consistent when you control device and network. Field data reflects your actual visitors, device mix and networks, so it often looks worse and changes slowly.

## What Lighthouse measures and how to run it

Lighthouse is an open-source lab tool in Chrome DevTools, the command line and PageSpeed Insights. It simulates a mid-range mobile under controlled CPU and network, then scores Performance and audits Accessibility, SEO and more. The Performance score is a weighted blend of metrics including Largest Contentful Paint, Cumulative Layout Shift and Interaction to Next Paint, as of 2026.

1. **Run in DevTools** Open your page, press F12, go to Lighthouse, choose Mobile, tick Performance, run. Close other tabs and extensions for cleaner results.
2. **Run from the CLI** Use the Lighthouse CLI to script tests in CI. Fix a throttling profile and device to keep runs comparable.
3. **Run multiple times** Run at least three times. Use the median to avoid an outlier from a network blip.

Use Lighthouse while you code. For example, change the hero image on /pricing, re-run, and see if LCP improves. A fixed page shows a faster LCP element and fewer long tasks in the trace.

## What PageSpeed Insights adds

PageSpeed Insights is a free web report. It runs Lighthouse like DevTools does. It also shows Chrome UX Report field data when available. That data is aggregated from opted-in Chrome users on your origin and URL patterns, over the last 28 days.

- Two panes: Field data at the top, then Lighthouse lab results below
- Field data shows Core Web Vitals distributions for your origin or page
- A pass or fail for Core Web Vitals over the last 28 days
- Real devices and networks from your audience, not a simulation

If your site is small, PageSpeed Insights may lack field data for some pages. It will still run Lighthouse. Treat that as lab-only and avoid over-generalising from one run.

## Why scores differ across runs and tools

- Different contexts: Lighthouse lab is one device and network; field data mixes many devices, networks and locations
- Different timeframes: the Lighthouse score is from one run; field data is a 28 day window
- Background noise: extensions, open tabs and CPU load can slow lab tests in a local browser
- Changes in the page: a new hero video on / guides/getting-started can drop LCP in field data a week after deploy
- Cache and CDN: a cold cache can hurt a lab test; field users often get warm cache hits

> Rule: compare like with like. Fix the same device, network and test location for lab. Compare field to field across dates, not field to lab.

## Which to trust for what

| Question | Use | Why |
| --- | --- | --- |
| What caused this slowdown on /pricing today? | Lighthouse | You need a repeatable trace and screenshots to see the bottleneck |
| Did last month’s changes improve real users’ core metrics? | PageSpeed Insights field data | You need the 28 day Chrome UX Report trend by metric |
| Does my page pass Core Web Vitals? | PageSpeed Insights field data | Pass or fail is based on real-user thresholds |
| What should I fix next? | Lighthouse diagnostics | Actionable audits and waterfall point to causes |
| Why does mobile feel slow on low-end devices? | Lighthouse mobile profile | You can throttle CPU and network for a realistic worst case |

For targets, use Core Web Vitals thresholds: LCP under 2.5 seconds, CLS under 0.1, INP under 200 ms. Aim to pass in field data. Treat a perfect Lighthouse score as nice to have, not the goal.

## Handling missing or sparse field data

Small sites often lack enough Chrome UX Report data at the URL level. Page or origin data may be missing or unstable. You still need to ship performance work.

- Optimise with Lighthouse and the Performance panel traces
- Test on a real mid-range phone over 3G or 4G, not just your desktop wifi
- Roll out to a few key templates first, then the rest
- Watch for field data to appear in PageSpeed Insights as traffic grows

When field data is missing, decide with your own timings too. Measure LCP and INP in your product analytics if you can instrument it. Keep claims modest until PageSpeed Insights shows a pass.

## A workflow you can run this week

1. **Pick one page** Choose /pricing or /guides/getting-started. It should be important and representative.
2. **Baseline in PageSpeed Insights** Record the field data status and the Lighthouse score. Screenshot the metric distributions.
3. **Debug in Lighthouse** Open DevTools, run Lighthouse and the Performance panel. Note the LCP element, long tasks over 50 ms and render-blocking requests.
4. **Fix the top bottleneck** For example, compress the hero image and set proper dimensions. Inline critical CSS. Defer a third-party script where safe.
5. **Re-test in Lighthouse** Run three times. Keep the median. Confirm the LCP element is faster and layout shift is stable.
6. **Deploy and wait** Field data lags. Give it time to roll through the 28 day window. Check PageSpeed Insights weekly.

A fixed /pricing usually shows an LCP down from 3.2 s to under 2.5 s in lab, a smaller CLS, and fewer long tasks in the main thread flame chart. Field data should follow if traffic and cache patterns match your tests.

## Reading the PageSpeed Insights field section

Start at the top. If it shows a Core Web Vitals assessment, that is your pass or fail based on real users over 28 days. Look at LCP, CLS and INP charts for the page and origin, if both are present.

- LCP: under 2.5 s is good. If many users are over, target images, render blocking and server time.
- CLS: under 0.1 is good. Fix shifts from images without sizes, web fonts and late ads.
- INP: under 200 ms is good. Reduce long tasks, hydrate less upfront, and avoid heavy input handlers.

Then map field problems to lab causes. If INP is poor in field, use Lighthouse traces to find long tasks and third-party scripts. If LCP is borderline, check the LCP element’s size, priority and whether it waits on render-blocking CSS or JS.

## Lab testing pitfalls that skew results

- Browser noise: disable extensions and close other tabs. Use Incognito.
- Emulation differences: DevTools and PageSpeed Insights both throttle, but exact conditions and locations differ.
- Network variance: wifi is bursty. Prefer the CLI in a clean CI runner for stable numbers.
- Cache state: run once with a primed cache to mimic repeat views, once cold for first visit.
- CPU: do not test on a high-end desktop only. Emulate a mobile profile and test on a real phone too.

If your Lighthouse score swings from 60 to 90 between runs, your environment is noisy. Stabilise it before you judge a change. A steady lab median is a better guide than a single high peak.

## When to bring in other testers

You can stay with Lighthouse and PageSpeed Insights for most work. If you need waterfalls and filmstrips from different locations and devices, run WebPageTest. If you want scheduled runs and extra visuals, GTmetrix adds that on top of Lighthouse. These tools are useful for debugging complex chains of requests and seeing the critical path clearly.

## The tools compared

### Lighthouse

Lab audits of a single page with controlled device and network. (developer.chrome.com)

Good at: Debugging render-blocking and long tasks; Identifying the LCP element and CLS shifts; Repeatable scoring across code changes; Running in DevTools and CI.

Choose it when You are coding and need immediate, repeatable feedback on a page before you ship.

Not for: Reporting Core Web Vitals from real users or judging origin-wide UX.

### PageSpeed Insights

Combining Lighthouse lab results with Chrome UX Report field data. (pagespeed.web.dev)

Good at: Core Web Vitals pass or fail from real users; Showing metric distributions over 28 days; Running Lighthouse without local setup; Comparing page versus origin field data.

Choose it when You want to know how real users experience a page and whether it passes Core Web Vitals.

Not for: Deep step-by-step debugging, scripting custom profiles or replacing full trace tooling.

**Verdict.** Use Lighthouse while you build to find and fix specific causes on a page under controlled conditions. Use PageSpeed Insights to see whether real users pass Core Web Vitals over the last 28 days and to communicate progress. When scores differ, trust PageSpeed Insights for impact and Lighthouse for diagnosis.

## Questions

### What is the difference between Google Lighthouse and PageSpeed Insights?

Lighthouse is a lab test you run in DevTools, the command line or inside PageSpeed Insights. PageSpeed Insights adds field data from the Chrome UX Report over the last 28 days and shows a Core Web Vitals pass or fail. Both show a Lighthouse lab score, but only PageSpeed Insights shows real-user data when available.

### Is Lighthouse performance reliable?

Yes, for lab checks, when you control the device, network and environment. Run multiple times and use the median. Disable extensions and background apps, or script runs in CI. Do not treat one high or low run as truth.

### What is considered a good Lighthouse score?

Use Lighthouse to find and fix causes, not to chase a number. A higher score usually means better lab performance, but your goal is to pass Core Web Vitals in field data: LCP under 2.5 s, CLS under 0.1, INP under 200 ms. Report the pass or fail from PageSpeed Insights.

### Is PageSpeed Insights accurate?

Its field section reflects real Chrome users on your site over the last 28 days, so it is the best view of user experience across devices and networks. It may not show field data for small sites or low traffic pages. Its lab section is a Lighthouse run, which is accurate as a controlled test when you stabilise your environment.

### Why does PageSpeed Insights not show field data for my page?

The Chrome UX Report needs enough real-user samples for that URL pattern or origin over the last 28 days. Small or new pages often fall short. Use Lighthouse to improve likely causes and check again as traffic grows. The origin-level view may appear before the page-level view.

### My lab score is green but field data fails. What should I do?

Your users have different devices, networks or usage than your lab emulation. Use Lighthouse traces to cut LCP and long tasks further. Reduce JavaScript, optimise images, prioritise the LCP element, and delay non-critical work. Test on a real mid-range phone over mobile data to align lab with reality.

## Read next

- [Core Web Vitals for a product site: the three numbers](https://porteur.ai/guides/core-web-vitals): The product founder’s guide to LCP, CLS and INP: what Google measures, why phones decide the pass, common causes, and how to test and fix.
- [Largest Contentful Paint: what slows it and what fixes it](https://porteur.ai/guides/largest-contentful-paint): See what usually counts as LCP, what makes it slow, how to fix it in order, and how to read it in Lighthouse and field data.
- [Reading a Lighthouse score on a phone](https://porteur.ai/guides/lighthouse-score): Use the mobile Lighthouse run to judge speed and UX. See what each score means, how lab and field data differ, and what to fix first.
- [Google Search Console vs Google Analytics: what each one knows](https://porteur.ai/compare/google-search-console-vs-google-analytics): Search Console shows pre click search data. Analytics shows what users do after the click. Why they differ, what to read in each, and how to join them.
- [Technical SEO checklist for a small site](https://porteur.ai/guides/technical-seo-checklist): Run these 20 technical SEO checks, in order. Each shows how to check it free and what fixed looks like for a site under 1,000 pages.
- [GTmetrix alternatives for page speed](https://porteur.ai/alternatives/gtmetrix-alternatives): Free ways to test your page on a phone with real user data, when to use each tool, and when paying for monitored tests is worth it.
- [PageSpeed Insights vs GTmetrix](https://porteur.ai/compare/pagespeed-insights-vs-gtmetrix): Both run Lighthouse. PageSpeed Insights shows real-user field data. GTmetrix adds waterfalls, locations, video and monitoring. Here is which to use first.
- [WebPageTest vs PageSpeed Insights](https://porteur.ai/compare/webpagetest-vs-pagespeed-insights): Use PageSpeed Insights for the verdict from field data, and WebPageTest to see why a page is slow. Here is when to choose each and how to pair them.

Paste your homepage URL and get a free check that reads your site and rivals on your searches, then shows three findings you can fix before you rerun Lighthouse. Free check: https://porteur.ai/