PageSpeed Insights: How to Read It and Improve Your Score
Short answer PageSpeed Insights is Google's free speed test. The top half shows what real Chrome visitors experienced over 28 days; trust it first. The bottom half is a one-off lab test that gives the 0–100 score and a list of suggestions. Fix the suggestions tied to a failing real-visitor metric, usually images, scripts and hosting, and ignore the rest.
Want help applying this to your business? Talk it through on a free 30-minute call.
Book a callThe short version: read the top half first
A PageSpeed Insights report has two parts that answer different questions.
| Part of the report | Where it comes from | What it tells you |
|---|---|---|
| Top: field data | Real Chrome visitors over the past 28 days | Whether your page actually feels fast to customers |
| Bottom: lab data | One simulated load on a mid-tier phone over a mobile network | The 0–100 score and a list of things to fix |
Source: PageSpeed Insights' own documentation. Google's web.dev team gives the rule for when the two disagree:
"As a general rule, if you have both field data and lab data for a given page, field data is what you should use to prioritize your efforts." (web.dev: Why lab and field data can be different)
So: if the top half passes, a yellow score below is not an emergency. If the top half fails, the bottom half tells you why.
How to run PageSpeed Insights
It's free and needs no account.
- Open pagespeed.web.dev and paste the full address of one page, such as your booking page, not just your home page.
- Press Analyze. Results come in a Mobile and a Desktop tab. Check mobile first: Google indexes and ranks the mobile version of your site (Google Search Central: mobile-first indexing).
- At the top of the field data, switch between This URL and Origin. Origin means every page on your site combined.
To see whole-site field data directly, type `origin:` before your address, for example `origin:https://www.example.com`. Suggestions only appear when you test a single page (PageSpeed Insights FAQ).
Field data: the Core Web Vitals assessment
This is the "Passed" or "Failed" line at the top. Core Web Vitals are three measures of how a page feels: how fast the main content shows, how fast it reacts to a tap, and how much the layout jumps.
| Metric | Measures | Good | Poor |
|---|---|---|---|
| LCP (Largest Contentful Paint) | Main image or text appears | 2.5 s or less | Over 4 s |
| INP (Interaction to Next Paint) | Page reacts to a tap or click | 200 ms or less | Over 500 ms |
| CLS (Cumulative Layout Shift) | Layout stays still while loading | 0.1 or less | Over 0.25 |
Thresholds: PageSpeed Insights documentation. The page passes when the 75th percentile of all three is good. In plain terms, at least three out of four visits must hit the target. If there isn't enough INP data, LCP and CLS alone decide it.
Below the three, you'll also see FCP (First Contentful Paint, when anything first appears) and TTFB (Time to First Byte, how fast your server answers). They don't decide pass or fail, but a slow TTFB usually points to hosting.
Under each metric, a green, amber and red bar shows the share of visits in each band. A bar that reads 11% amber means 11% of visits fell between good and poor.
Old guides list FID (First Input Delay) here. INP replaced it on March 12, 2024 (web.dev).
"No data" in the top half
A newer or low-traffic page often has too few Chrome visitors. PageSpeed Insights then falls back to your whole site, and if the site is also too small, it shows no real-visitor data at all (PageSpeed Insights documentation). That's common for local businesses and isn't a penalty. Use the lab section instead, and check Search Console's Core Web Vitals report over time (how to read Search Console).
Lab data: what the 0–100 score means
The score comes from one lab run. 90–100 is good, 50–89 needs improvement, under 50 is poor. It's a weighted average of five lab measurements:
| Lab metric | Weight |
|---|---|
| Total Blocking Time (how long scripts freeze the page) | 30% |
| Largest Contentful Paint | 25% |
| Cumulative Layout Shift | 25% |
| First Contentful Paint | 10% |
| Speed Index (how quickly the screen fills in) | 10% |
Weights: Chrome for Developers, Lighthouse performance scoring. Lighthouse is the testing engine inside PageSpeed Insights.
Three things owners get wrong about the score:
- It moves between runs. Network, server load and test hardware all vary, so the same page can score 62 then 70 with no change (PageSpeed Insights documentation). Run it three times and look at the middle result.
- Suggestions don't add points. Google says "only metrics contribute to your Lighthouse Performance score," not the suggestions themselves. Fixing one helps only if it moves a metric.
- Google ranks on the field data, not this number. Google says Core Web Vitals "are used by our ranking systems," and that chasing "a perfect score just for SEO reasons may not be the best use of your time" (Google Search Central: page experience). More on ranking: does website speed affect Google ranking.
The lab can't measure INP, because nobody taps the page during the test. Total Blocking Time is its stand-in (web.dev).
What each suggestion means and what to do
Since October 20, 2025, PageSpeed Insights runs Lighthouse 13, which replaced the old "Opportunities" list with Insights (PageSpeed Insights release notes; What's new in Lighthouse 13). Old advice uses old names; the table maps both.
| Insight (old name) | What it means | What to do |
|---|---|---|
| Improve image delivery (Serve images in modern formats, Properly size images) | Photos are bigger than the screen needs | Resize to display size, compress, save as WebP or AVIF. Images saving under 4 KiB aren't flagged |
| LCP request discovery | The main photo starts loading late | Don't lazy-load the top image; mark it high priority (`fetchpriority=high`) |
| LCP breakdown | Shows where the main-content delay sits | Read it to see whether the server, the image or the code is slow |
| Render-blocking requests (Eliminate render-blocking resources) | CSS or scripts hold up the first paint | Defer what the first screen doesn't need; on WordPress, use a speed plugin |
| Document request latency (Reduce server response times) | Your HTML arrives late: redirects, a server slower than 600 ms, or no compression | Remove redirects, turn on compression, add caching or upgrade hosting |
| Layout shift culprits | Names the elements that jump | Set width and height on images and video, reserve space for banners |
| Use efficient cache lifetimes | Repeat visitors re-download files | Cache images, fonts and scripts for at least 30 days |
| Font display | Text stays invisible while a custom font loads | Set `font-display` to swap or optional |
| Third parties | Lists chat, tracking and booking scripts | Informational only; remove tools you no longer use |
Sources: Chrome for Developers' page for each insight. Below the Insights list, a Diagnostics list (for example, "Reduce unused JavaScript") keeps the audits that didn't change (Chrome for Developers: Moving Lighthouse to insights).
Read the table from the failing field metric. Failing LCP: images, LCP discovery, render-blocking, server latency. Failing CLS: layout shift culprits, font display. Failing INP: third-party and other heavy scripts.
On Wix, Squarespace or Shopify, the platform runs the server and code. You control image size, embeds and apps (platform choice and SEO).
Which suggestions to skip
Not every line is worth the work. Google says:
"You have to consider the tradeoffs for your own application: some of PageSpeed's recommendations take substantial effort to implement, so you need to evaluate the cost of making changes vs. the benefit the rule would have on your page." (PageSpeed Insights FAQ)
Skip a suggestion when the field data already passes on that metric, the saving is tiny, or the fix means dropping a tool that brings in leads, like your booking widget. "Defer offscreen images" no longer exists: Google removed it because lazy-loading below-the-fold images rarely changes what Lighthouse measures (What's new in Lighthouse 13).
After you fix: when the numbers change
The lab score updates the moment you rerun the test. The field data covers the trailing 28 days, so a fix shows up gradually and fully only about four weeks later (PageSpeed Insights documentation).
Example
Illustrative cases, not measured results.
A landscaper with a mobile score of 41
Field data: passed. The score comes from a 3 MB slideshow flagged under image delivery. He compresses the photos to WebP; the score rises to the 70s. The real-visitor result was already fine, so he stops there.
A dental office failing LCP at 3.8 seconds
LCP request discovery shows the top photo is lazy-loaded, and document request latency shows two redirects. Removing the lazy-load and pointing links straight at the final address costs an afternoon. Four weeks later, field LCP reads 2.3 seconds.
A new bookkeeping firm with no field data
Too few visitors yet. She runs the lab test three times on her contact page, fixes the CLS culprit (a cookie banner without reserved space) and checks again in a few months.
FAQ
Is PageSpeed Insights free?
Yes. The website is free with no sign-in. Developers can also run it through an API.
Why is my mobile score so much lower than desktop?
Mobile simulates a mid-tier phone on a mobile network; desktop simulates a wired computer (PageSpeed Insights documentation). Weaker hardware means slower loads.
Do I need a score of 100?
No. 90 counts as good, and Google says good scores don't guarantee top rankings (Google Search Central: page experience). Aim for passing field data.
Does the SEO score in PageSpeed Insights measure my rankings?
No. It's a basic Lighthouse checklist, run in the same lab test. Rankings depend on far more.
How often should I test?
After any site change, and monthly otherwise. Field data moves on a 28-day window, so weekly checks add little.
Different for you?
- Whether speed is hurting your rankings at all: Core Web Vitals and Google ranking.
- Field data for the whole site, grouped by page type: Google Search Console for a small business.
- Your builder limits what you can fix: Wix, Squarespace or WordPress for SEO.
- Fast page, still no calls: website not getting leads.
- Want the site made fast and kept fast: Website Agent.