How Google Measures Page Speed in 2026

Almost every WordPress owner has had the same confusing moment: Lighthouse hands you a green 95, you feel great, and then Google Search Console quietly tells you the same page is "failing" Core Web Vitals. Both tools come from Google. Both claim to measure speed. So which one is lying?

Neither. They measure two completely different things, and understanding the difference is the entire key to understanding how Google actually measures the speed of your site in 2026. This article is about measurement — what Google records, where the numbers come from, and how they turn into a pass or fail — not about how to tune your theme. Once you know what the scoreboard is really tracking, the tuning makes a lot more sense.

Google measures real visitors, not your test runs

The number that matters for ranking is not a score from a tool you ran. It comes from the Chrome User Experience Report, usually called CrUX. Every time a real person visits your site in Chrome with usage statistics enabled, their browser can report back how that specific page performed on their specific device and network. Google aggregates those millions of real sessions into a public dataset.

This is called field data, and it is the only speed data Google uses for ranking. It reflects a grandmother on rural DSL, a commuter on a flaky 4G train, and someone on fibre with a flagship phone — all blended together. You cannot fake it, and you cannot improve it by running a faster test. You can only improve it by genuinely shipping a faster experience to actual humans.

The 28-day rolling window and the 75th percentile

Two details about CrUX trip people up constantly.

First, the field data Search Console and PageSpeed Insights show you is a 28-day rolling average. If you deployed a major fix today, your field scores will not flip to green today. They will improve gradually as the slow days from before your fix age out of the window. Give it three to four weeks before you judge whether a change worked.

Second, Google does not grade you on your average visitor. It grades you on the 75th percentile of your visitors. That means the slowest 25 percent of your page loads have to meet the threshold before the page counts as "good." This is deliberately strict: it forces you to care about your worst-served users, not just the lucky ones on fast hardware. A site that feels fast on your MacBook can still fail because a quarter of its real traffic is on cheap Android phones over congested mobile networks.

The three metrics that actually count

As of 2026 there are exactly three Core Web Vitals, and each one targets a different felt frustration:

• Largest Contentful Paint (LCP) — how long until the biggest visible element (usually your hero image or headline) finishes rendering. Good is 2.5 seconds or less. This is your "is anything happening yet?" metric.
• Interaction to Next Paint (INP) — how quickly the page visibly responds when someone taps or clicks, measured across the whole visit. Good is 200 milliseconds or less. This is your "did that button do anything?" metric.
• Cumulative Layout Shift (CLS) — how much the page jumps around as it loads. Good is 0.1 or less. This is the "I tapped the wrong thing because the layout moved" metric.

One critical update many older guides get wrong: INP replaced First Input Delay (FID) in March 2024. FID is retired and gone. If a plugin's marketing or a tutorial still talks about FID as a current metric, it is out of date, and you should treat the rest of its advice with suspicion. INP is harder to pass than FID ever was, because it measures every interaction during a visit rather than just the first one — which is exactly where heavy WordPress sites loaded with page builders and third-party scripts tend to stumble.

What is NOT a Core Web Vital

Time to First Byte (TTFB), First Contentful Paint (FCP), and Speed Index show up all over performance tools and get talked about as if they were ranking metrics. They are not. They are diagnostics — useful for figuring out why your LCP is slow, but Google does not rank on them directly. TTFB in particular matters enormously as a cause (slow shared hosting and unoptimised PHP wreck it), but it is a lever, not the scoreboard. Don't let a tool that obsesses over TTFB convince you it's a vital.

Lab data versus field data — why your scores disagree

Now we can solve the opening mystery. When you run a tool like Lighthouse (the engine inside the Chrome DevTools audit and the top half of PageSpeed Insights), you get lab data: a single simulated load on a throttled connection and a mid-tier emulated phone, in a clean environment with no cookies, no logged-in state, and no real-world chaos. It produces that famous 0-to-100 performance score.

That score is a useful lab signal, but Google does not use the Lighthouse score for ranking at all. It is a synthetic prediction, not a measurement of your users. So you can absolutely score 95 in the lab while your 75th-percentile real users fail in the field — because your real visitors carry slower hardware, ad scripts, consent banners, and tracking tags that a clean lab run never sees. The reverse happens too: a modest lab score with great field data, common on lean sites with loyal repeat visitors who benefit from caching.

On PageSpeed Insights, the rule of thumb is simple: look at the field data at the top first. If that panel exists and is green, you are passing in Google's eyes regardless of what the lab score below says. The lab section is only there to help you diagnose. When the two disagree, the field data is the truth and the lab data is the hypothesis.

Why small WordPress sites often see "no data"

Here is a quiet frustration for new sites: CrUX only reports data for pages and origins that get enough Chrome traffic to be statistically meaningful and anonymous. A brand-new blog or a low-traffic local business site frequently has no page-level field data at all. PageSpeed Insights will fall back to origin-level data (your whole domain averaged together) or show nothing and lean entirely on the lab score.

This is normal, not a bug. It also means a freshly launched site genuinely has no Core Web Vitals ranking signal yet — there's nothing to measure. Until you have traffic, the Lighthouse lab score and a tool like WebPageTest are your best proxies. Once real visitors arrive, the field data takes over as the only thing that counts.

Where to read your real numbers

For a WordPress owner, three free sources cover everything:

1. Search Console's Core Web Vitals report groups every URL on your site into "Good," "Needs improvement," and "Poor," split by mobile and desktop, straight from CrUX. This is the closest thing to Google's own gradebook for your domain, and it's where you should check first.
2. PageSpeed Insights gives you a single URL's field data plus a full lab diagnosis in one place. Use it to investigate a specific page Search Console flagged.
3. Google Site Kit, the official WordPress plugin, pulls PageSpeed Insights and Search Console data into your wp-admin dashboard so you don't have to leave the site to glance at your standing.

For pages with no field data, WebPageTest lets you run repeatable lab tests from a fixed location, browser, and connection — the controlled before-and-after numbers you need when CrUX can't see you yet.

How much does this actually move rankings?

Be realistic. Page experience, of which Core Web Vitals is the measurable core, is a real Google ranking signal — but a lightweight tiebreaker, not a heavyweight factor. Google has been consistent on this: relevant, helpful content beats fast content nearly every time. A blazing-fast page about the wrong thing will not outrank a slightly slower page that genuinely answers the query.

Where speed earns its keep is at the margins. When two pages are comparably useful, the faster one wins the tiebreak. And the field-data nature of the signal means the real payoff is human, not algorithmic: passing Core Web Vitals at the 75th percentile means your slowest quarter of visitors are still having a tolerable experience — which protects conversions, bounce rate, and the repeat visits that actually grow a site. Chase the green because your real users feel it, and let the small ranking nudge be the bonus.

Measure the field, not the lab. Wait out the 28-day window. Grade yourself at the 75th percentile, on mobile, like Google does. Do that, and the gap between your green Lighthouse score and your Search Console verdict stops being a mystery — and starts being a map.