You run your homepage through PageSpeed Insights and get a score of 58. Alarming. Then you check your Core Web Vitals report and everything is green. Or the other way round: a score of 96, but customers keep telling you the site feels slow.

Neither result is wrong. They are measuring two different things.

The short answer: a PageSpeed score is a lab test run on a simulated device. Core Web Vitals from real visitors are field data. Use the lab test to find out why a page might be slow, and the real-visitor data to decide whether it is a problem.

What a PageSpeed score actually is

PageSpeed Insights runs Lighthouse, an automated test, against your page on Google's servers. It loads the page once, on a simulated device and connection, and scores it from 0 to 100.

That makes it consistent and repeatable, which is useful. But it is still a single simulated visit:

It does not know what devices your visitors actually use.
It does not know how fast their connections are.
It loads the page once, so results can vary by several points between runs.
It cannot tell you what happens after the page loads, such as a visitor tapping a menu.

What real-visitor data is

Real-user data is collected from the people who actually visit your site, on their own phones and laptops, over their own connections. This is what Core Web Vitals are based on, and it is the experience your visitors have.

It shows the spread of experiences, from the visitor on fast office wifi to the one on a patchy mobile signal, which no single lab run can show. It needs traffic to build up, so low-traffic pages can have limited data.

Why they disagree

Common reasons a lab score and real-visitor data tell different stories:

Your visitors are not the test device. If most visitors use fast, recent phones, real results can be better than a throttled lab test. If many are on older devices, they can be worse.
Caching helps real visitors. Returning visitors often have images and scripts already stored. A lab test starts from scratch.
Lab tests miss real interactions. Responsiveness to taps and clicks depends on what visitors actually do.
Third-party scripts behave differently. Chat widgets, ads and consent tools can load differently in real life than in a test.
Run-to-run noise. The same page can score a few points differently on consecutive runs.

So which one should you trust?

Treat them as two halves of one answer:

	Lab score (Lighthouse)	Real-visitor data (Core Web Vitals)
Answers	Why might this page be slow?	Is this page slow for people?
Strength	Specific, fixable causes	Reflects real experience
Weakness	One simulated visit	Needs enough traffic
Best used for	Finding what to fix	Deciding if it matters

If real-visitor data passes, a mediocre lab score is not an emergency. It may still point at easy wins, such as an oversized image. If real-visitor data fails, the lab diagnostics are your best clue for why.

What to do with this

Check real-visitor data first. Is the page passing or failing for actual people?
Then read the lab diagnostics. Look for the specific causes, such as oversized images or heavy scripts, rather than chasing the score.
Compare against last time, not a perfect 100. A page that moved from 62 to 75 has improved, even if it is not at 90.
Fix one thing, then re-check. Changes show up in real-visitor data over the following weeks.

For more on the metrics themselves, see our plain-English guide to Core Web Vitals, and our monthly website performance checklist.

How we use both

Web Vitals Collector measures real visitors on your site and treats that as the authority on whether a page passes or fails. We then use Google's PageSpeed Insights to explain the likely causes, and compare the findings with the previous run so you can see what changed. Both are combined into one plain-English weekly report, with the most useful fixes first.

You can see what that looks like in our sample report.

PageSpeed and Lighthouse results are automated estimates and are not a guarantee of search ranking.