Lighthouse vs. PageSpeed Insights: A Score of 100 Proves Nothing

A Lighthouse score of 100 gets waved around as proof that a website performs well. It proves exactly one thing: that a single simulated test run looked good under lab conditions. What real visitors on real devices with real connections actually experience is a different question entirely - and the answer to that question is not Lighthouse, it is field data.

Why a Lighthouse score of 100 proves nothing

Lighthouse runs a single, controlled test: a simulated device, a simulated network speed, no variance from real users. That makes the number reproducible and well suited to quickly checking, during development, whether a change moves things in the right direction. It also makes it unsuited as sole proof of real-world performance, because no real visitor loads the page under exactly these lab conditions. A score of 100 on a powerful test machine with a stable connection says nothing about how the page feels on an older mid-range phone on a mobile network - which is exactly where most sites actually lose visitors.

Lab data: what Lighthouse actually measures

Lighthouse is a lab test: a fixed, repeatable test environment, available directly in Chrome DevTools or as a standalone command-line tool. It scores performance, accessibility, SEO fundamentals and best practices each from 0 to 100 and pairs the score with a detailed list of concrete recommendations. Its big advantage is speed: a run takes seconds and can be repeated as often as needed without waiting for real traffic - ideal for confirming during development that a change actually helps.

Field data: what PageSpeed Insights adds

PageSpeed Insights combines that same lab test with a second, more important data source: the Chrome User Experience Report (CrUX), an aggregated dataset built from the actual loading experiences of real Chrome users who visited a page. This field data reflects real device diversity, real network conditions and real user behaviour - and it is this field data, not the lab score, that feeds Google's page experience signal. We cover what other signals feed into that assessment alongside Core Web Vitals in our article on Page Experience 2026. Google Search Console shows the same CrUX data aggregated by URL group under Page Experience.

When field data is missing, and what that means

CrUX only publishes data for URLs that have collected enough measurements within a rolling 28-day window. New pages or low-traffic pages therefore show only the lab portion in PageSpeed Insights, with no field data - that is not a misconfiguration, it is a matter of time and traffic. If you need certainty about real user experience during that period, add your own real-user monitoring, which measures independently of the public CrUX dataset and needs no minimum visitor count.

Concrete steps for the next 90 days

• Days 1-30: run Lighthouse in Chrome DevTools on your most important pages and fix the obvious lab-level issues
• Days 31-60: check PageSpeed Insights for the same URLs, compare field data against lab values, and document any gap
• Days 61-90: where field data is missing, set up real-user monitoring, then check both tools together every month

Conclusion

Lighthouse is the workshop tool for fast iteration, PageSpeed Insights is the reality check that follows. Using both for what they are actually good at, instead of presenting a lab score as proof on its own, is what optimising for what actually matters looks like. The current thresholds for LCP, INP and CLS are covered in our article on Core Web Vitals 2026. How a modern rendering architecture affects those numbers is covered in our article on JavaScript SEO for SPAs. What a technical performance optimisation typically costs is covered in our article on SEO costs. For a no-obligation assessment of your current lab and field values, feel free to get in touch.