I used to think Google changed my site's ranking based on Lighthouse scores it would measure every time their bot visits my site. 🚦 I was so WRONG! 🤦♂️ It turns out that Chrome browsers quietly gather Core Web Vitals from real users, and Google uses the 75th percentile of this data from the past 28 days to adjust rankings. In other words, 100 points Lighthouse is not your ultimate goal. Perfect in-field measurements on real users' browsers is what you're looking for.
Disclaimer: This article refers to PageSpeed Insights v5 and Lighthouse 12.4.0. When you're reading it, this info might be outdated due to newer versions
When you use Google PageSpeed Insights, you'll see two sections. One is with real users' Web Vitals data with horizontal bars (not a circle) focus only on that, look no further!
See an example report for the website https://web.dev:
The second section is not important for Google's ranking at all. It merely reflects the outcome of an on-demand Lighthouse simulation test that was just run on demand on a Google machine. This is just an artificial simulation, susceptible to many variable factors, and it can sometimes be misleading if you're unaware of the complex processes it performs behind the scenes — knowledge that, unfortunately, is not commonly shared.
Therefore, the Lighthouse test serves only as a "compass," not an exact "pin on a map" indicating your desired destination. It provides rough directions on where to go and what to improve. Additionally, the Lighthouse test is an artificial tool, much like a "compass," and is highly susceptible to "local magnetic field disturbances" (e.g., temporary poor network conditions). Its non-deterministic nature arises from the algorithm calculating two artificial scenarios—optimistic and pessimistic network call dependency graphs—and averaging them, which is sensitive to race condition edge cases based on your actual recorded network call dependency graph.
