Quick answer: Google PageSpeed Insights gives your URL a score from 0 to 100 for mobile and desktop, based on a Lighthouse lab test. For many sites it also shows field data from real Chrome users when enough traffic exists. Treat the score as a health check, not a trophy. Read the opportunities list, fix the biggest issues first (images, caching, unused scripts) and retest the same pages after each change. A mid-70s score with a fast feel for users beats a brittle 99 that breaks features.
What PageSpeed Insights actually runs
When you paste a URL into PageSpeed Insights, Google runs Lighthouse against that page. Lighthouse loads the page in a controlled environment, collects timings and applies a weighted formula to produce the 0-100 score. Mobile and desktop are separate runs. The mobile test uses a simulated slower device and throttled network, which is why mobile scores are almost always lower.
Alongside the lab run, PageSpeed Insights may show a field data section drawn from the Chrome User Experience Report. That section only appears when Google has enough anonymised data for your origin or URL. Field data is closer to what real visitors experience. Lab data is better for debugging because it is repeatable under known conditions.
For a wider view of other speed tools, see our comparison of website speed test options.
How to read the score bands
Google groups scores roughly as:
- 90-100: Good
- 50-89: Needs improvement
- 0-49: Poor
Small swings of a few points between runs are normal. Network noise, third party scripts and ads can move the number. Do not panic over a drop from 84 to 79 after adding a chat widget. Panic if the page feels slow and opportunities show multi-second savings.
Lab metrics you will see in early 2020
Lighthouse reports several timings. Names and weights change over time, but these ideas stay useful:
- First Contentful Paint (FCP): When the first text or image appears. Visitors need something on screen quickly.
- Speed Index: How quickly content becomes visually filled in.
- Time to Interactive (TTI): When the page is reliably ready for input. Heavy JavaScript pushes this out.
- First Meaningful Paint / Largest content metrics: Roughly when the main content appears (wording in reports has shifted as Lighthouse evolved).
- Total Blocking Time / Max Potential FID style measures: How much main-thread work blocks interaction.
You do not need to memorise every label. Ask: does content appear quickly, and can I tap without waiting for scripts to finish?
Opportunities and diagnostics
The value of PageSpeed Insights is the opportunity list, not the vanity score. Common items for WordPress and marketing sites:
- Serve images in next-gen formats and correctly sized dimensions
- Eliminate render-blocking CSS and JavaScript
- Remove unused CSS and JavaScript
- Enable text compression
- Serve static assets with an efficient cache policy
- Minify CSS and JavaScript
- Reduce server response times (TTFB)
- Preconnect to required origins
- Defer offscreen images (lazy loading)
Each opportunity estimates time saved. Sort by estimated savings and fix the largest first. Our guides on website caching, minifying CSS and JavaScript and WordPress image optimization cover the usual wins.
Mobile vs desktop: why the gap
Mobile scores hurt because the test assumes a weaker CPU and a slower connection. Desktop scores can look fine while phones struggle. Always check the mobile tab for public marketing pages. If your audience is mostly on phones (local services, retail, many B2C sites), mobile is the report that matters.
Also test on a real phone over cellular data. A lab score is a guide; your thumb is the judge.
Field data vs lab data
When field data appears, compare it to the lab score. Sometimes lab looks poor because of a one-off heavy third party script during the test, while real users on faster connections are fine. Sometimes lab looks okay but field data shows slow loads in a region far from your server. That points to CDN or hosting location issues.
If field data is missing, your traffic may be too low for the report. Keep using lab tests and real device checks.
A practical workflow
- Pick 3-5 important URLs: homepage, main service or product page, blog post with media, checkout or contact.
- Run PageSpeed Insights on mobile for each. Save the report links or screenshots of opportunities.
- Fix the top two opportunities that apply sitewide (often images and caching).
- Retest. Confirm the page still works: forms, menus, bookings, analytics.
- Tackle page-specific issues (a huge hero video, an embed on one landing page).
- Recheck monthly and after plugin or theme changes.
Pair this with WordPress speed optimization if your site runs on WordPress, and with mobile page speed habits for phones.
What not to do
- Do not strip every marketing pixel just to hit 100 if those tools pay for themselves
- Do not install five optimisation plugins that fight each other
- Do not ignore server quality; no plugin fixes chronic oversold hosting
- Do not chase micro-optimisations on pages nobody visits
- Do not change hosting or themes based on a single score alone
Third party scripts and the score trap
Marketing sites often load analytics, tag managers, chat, heatmaps, A/B testing and ad pixels. Each one can add CPU work and network delay. PageSpeed Insights will list many of them under third party opportunities.
Before you delete everything, ask which tools earn their keep. Analytics that you actually read are worth a small cost. A chat widget nobody answers is not. Load heavy tags on thank-you pages only when that is enough. Delay non-critical scripts until after first interaction when your developer can do it safely.
Also watch tag managers that inject surprises. A tidy GTM container with unused tags still firing will quietly punish every page.
Hosting and time to first byte
If the server responds slowly, front-end tips only go so far. Shared hosting under load, distant data centres and uncached dynamic pages raise TTFB. Managed WordPress hosts with page caching, or a well configured cache plugin plus a CDN for static files, usually help more than micro-optimising CSS selectors.
Test from a location near your audience when you use tools that allow it. PageSpeed Insights itself does not let you pick a city, so pair it with WebPageTest or GTmetrix when geography matters.
Images: the usual biggest win
Oversized hero images remain the top offender on business sites. Export at the display size you need, compress, and prefer modern formats when your audience browsers support them and your stack can serve them. Use srcset for responsive images. Lazy load below-the-fold media. Do not lazy load the main above-the-fold hero if that delays the first big image; balance is required.
Background videos autoplaying on mobile destroy scores and patience. Use a static poster and optional play instead.
CSS and JavaScript weight
Page builders and multipurpose themes often enqueue assets globally. Ask your developer to load builder CSS only where used, defer non-critical JavaScript and remove unused plugins. Critical CSS techniques help but are easy to break; start with fewer plugins and a lighter theme before complex critical-path tooling.
Minification helps a little. Removing unused code helps a lot.
Retesting discipline
Change one major factor at a time when you can. Retest mobile on the same URLs. Record scores in a sheet with date and what changed. Ignore tiny fluctuations. Celebrate when opportunities that showed multi-second savings disappear and when real users stop complaining about lag.
FAQ
Is PageSpeed Insights the same as Search Console speed reports?
Related but not identical. PageSpeed Insights is a URL lab tool with optional CrUX field data. Search Console has had speed-related reports that draw on Chrome data at site level. Use both when available.
Does the score affect rankings?
Google has used page speed as a ranking signal for years, more strongly for the slowest experiences. Relevance still dominates. Speed mainly helps when two results are otherwise similar, and it always helps conversion.
Why did my score drop after I added a booking widget?
Third party scripts cost CPU and network time. Weigh the lead value against the speed cost. Load widgets only on pages that need them when you can.
Can I run Lighthouse locally instead?
Yes. Chrome DevTools includes Audits (Lighthouse). Useful for staging sites and logged-in pages that public PageSpeed Insights cannot reach.
Use the report, then move on
PageSpeed Insights is a flashlight, not a finish line. Read the opportunities, fix what hurts users, and keep an eye on real visits. If you want a site that loads quickly and converts, see our Turn visitors into customers service.
