Google PageSpeed Insights vs Lighthouse: Key Differences

“My PageSpeed score is 78 but my Lighthouse score is 65. Which one should I trust?”

This is a common source of confusion. PageSpeed Insights and Lighthouse are closely related but not identical. Here’s what’s actually happening.

What They Share

Both tools use the Lighthouse engine to generate performance scores. The metrics they measure — LCP, CLS, TBT, FCP, Speed Index — are the same, as are the thresholds for good/needs improvement/poor.

Lighthouse is the open-source auditing tool. PageSpeed Insights is Google’s web interface for running Lighthouse in a standardized environment plus adding CrUX field data.

What Makes Them Different

The Testing Environment

Lighthouse CLI / Chrome DevTools: Runs on your local machine with your CPU, your network conditions, your installed extensions. Results vary based on your hardware and what else is running. A powerful MacBook Pro will produce much better scores than the throttled conditions Lighthouse uses by default.

PageSpeed Insights: Runs in Google’s data centers with standardized, reproducible conditions. This makes scores more consistent run-to-run.

debugutils.dev: Also runs on dedicated server infrastructure with controlled conditions — no extensions, fresh cache, consistent throttling. More reproducible than local Lighthouse.

CrUX Field Data

This is the biggest difference. PageSpeed Insights pulls in CrUX (Chrome User Experience Report) data — real performance metrics from actual Chrome users who visited your site. This data is:

• What Google uses for ranking
• The 75th percentile of real user experience
• Aggregated over 28 days

Lighthouse (local or via debugutils.dev) only gives you lab data — a single simulated test run. Field data often differs from lab data because real users are on varying devices, networks, and conditions.

If your site has enough traffic, PageSpeed Insights shows both lab and field data. For sites with low traffic, only lab data is available.

Score Differences

Even the lab scores from PageSpeed Insights vs local Lighthouse often differ. Reasons:

• Different CPU throttling — PageSpeed Insights uses a specific simulated CPU slowdown that may differ from your local Lighthouse settings
• Different network throttling — Same issue; the throttling profiles don’t always match exactly
• Server-side rendering variability — Dynamic pages may serve different HTML in different environments
• Chrome version — Lighthouse versions get updated; older local installations may use different weights

A 5–15 point difference between local Lighthouse and PageSpeed Insights is normal. For another angle on tool differences, see GTmetrix vs PageSpeed Insights.

Which One Should You Trust?

For ranking/SEO purposes: PageSpeed Insights field data. This is what Google sees.

For diagnosing specific issues: Use a lab tool with diagnostic depth — debugutils.dev for waterfall + filmstrip + LCP analysis, or Lighthouse DevTools for quick local checks.

For CI/CD performance gates: Lighthouse CLI or a server-side tool with consistent conditions. Local Lighthouse is too noisy for pass/fail decisions.

Practical Workflow

1. Establish baseline: Run PageSpeed Insights to see your CrUX field data and understand where you stand with real users.

2. Diagnose issues: Run debugutils.dev to see the waterfall, LCP element, filmstrip, and Lighthouse recommendations in one report.

3. Make changes, verify with lab data: Run debugutils.dev before/after changes to confirm improvements.

4. Monitor field data improvement: Check PageSpeed Insights 2–4 weeks after deploying changes to see field data update.

The two tools complement each other. Neither alone gives you the complete picture. For a full rundown of available options, see Best Website Speed Test Tools.