When using PageSpeed Insights to improve a web page, the key is to check the mobile and desktop reports separately, choose fixes based on specific diagnostics, then review the metrics and try using the page after making changes, rather than simply chasing an overall score. It can help identify problems with loading, interaction, and layout stability. But simulated tests do not represent every visitor’s experience, and a higher score does not necessarily mean the site has actually become faster.
Understand the data sources before reading the three core metrics
Enter the URL of the page you want to check and start the analysis. Review the mobile and desktop results separately, then scroll down and expand the diagnostic items to read their explanations and recommendations. The report’s field data comes from the Chrome User Experience Report (CrUX) and reflects real-world usage. Its lab data comes from Lighthouse’s simulated tests in a controlled environment. The two use different measurement methods.
- LCP: Measures how long it takes for the largest content element in the viewport to appear, helping you assess whether users are waiting too long for the main content.
- INP: Measures responsiveness to interactions, focusing on whether users have to wait after taking an action.
- CLS: Measures visual stability, helping you identify whether elements shifting during loading interfere with use.
For example, text appearing on the screen does not mean the main content is fully visible. FCP and LCP answer different questions, so you should not focus solely on how quickly the first content appears.
Turn diagnostics into an action list instead of applying every fix at once
You can prioritize problems that affect most visitors or start with situations where the experience is worst, then plan the work around the report’s diagnostics. A low score alone is not a reason to immediately switch hosting providers or install a plugin.
- Oversized images and resources: Check image file sizes and consider whether compression is needed. If the report points to code resources, review the complexity and file sizes of CSS and JavaScript.
- Too much code and too many external calls: Share the relevant diagnostics with technical staff, identify which resources need to stay, then assess what can be reduced or removed.
- Slowdowns after a recent update: Check whether newly deployed technology is affecting TTFB—the time from sending a request to receiving the first byte from the server.
Choose these approaches according to the problem; they are not a fixed checklist for every website. For changes involving code or servers, ask technical staff to help assess what is appropriate.
Retest after making changes, and account for differences between test versions
After applying fixes, check the same page’s mobile and desktop reports again, compare the metrics that originally showed problems, and test whether the layout, buttons, and links are easy to use. If you suspect a recent change has slowed the site down, consider reverting it. Then monitor whether performance improves to check whether your suspicion was correct.
Keep tool updates in mind when comparing scores. Search Engine Journal reported in November 2026 that Lighthouse 13.6 changed how FCP and LCP are simulated. Scores may therefore change even if the page itself has not. A single increase in the overall score is not enough to declare an optimization successful. If you are also checking accessibility, a tool’s score cannot replace a full WCAG audit and manual checks.
Improving web pages with PageSpeed Insights: perspectives from across the web
If no field data is displayed, does that mean the page has no problems?
You cannot draw that conclusion. A 2026 report explains that field data reflects Chrome users’ actual experiences, while a 2019 article introducing version 5 of the tool says that data may not be displayed when there is too little of it. Missing field data does not mean performance is good. You can still consult the lab diagnostics, while keeping in mind that they are simulated results and do not directly represent every visitor’s experience.
Where can you start when images cause layout shifts?
Start by checking whether image elements have explicit width and height attributes, and pay attention to any large layout shifts identified in the report. In its version 16.0 integration update, Screaming Frog listed both as diagnostic areas that can help improve CLS. You still need to confirm what is causing shifts on the specific page rather than assuming that images cause every CLS problem.
Do worse simulated results necessarily mean the real-world experience has also worsened?
Not necessarily. Search Engine Journal reported on a Framer test page where preloading JavaScript modules improved real-world time to interactivity, but produced a worse LCP result in the older simulation. This was a specific test case, not evidence that preloading always helps. If simulated results and actual performance point in opposite directions, assess them alongside real-user data and how the page behaves in use.
References
- Screaming Frog SEO Spider Update – Version 16.0www.screamingfrog.co.uk · 2021-09-22T07:45:08+00:00 — Discusses: Where can you start when images cause layout shifts?
- 7-Step Ecommerce Website Audit to Boost Conversionswww.semrush.com · 2024-05-14T08:30:00Z — Source used in the article above.
- 【2026年SEO軟體推薦】30款SEO工具+挑選指南www.yesharris.com · 2026-03-27T16:14:24+08:00 — Source used in the article above.
- Google Lighthouse Update Changes How PageSpeed Insights Simulates Loadingwww.searchenginejournal.com · 2026-11-02T21:52:57+00:00 — Discusses: Do worse simulated results necessarily mean the real-world experience has also worsened?
- PageSpeed是什麼?會影響SEO嗎?網站測速工具這裡看!welly.tw · 2024-10-07T10:41:12.274Z — Source used in the article above.
- 「宅網站」讓你的網站關鍵字排不上去www.da-vinci.com.tw · 2020-12-02T23:46:11+08:00 — Source used in the article above.
- Small Business SEO: 10 Tips to Get Your Website Found Onlinebacklinko.com · 2024-06-17T09:13:21-05:00 — Source used in the article above.
- 19 Ways to Diagnose a Decline in Organic Trafficwww.seerinteractive.com · 2026-05-26T00:00:00-05:00 — Source used in the article above.
- Google 為什麼不喜歡緩慢的網站?tenten.co · 2023-12-12T19:07:24.000Z — Source used in the article above.
- 認識SEO的排名因素:Page Experience 網頁體驗訊號www.yesharris.com · 2023-10-03T20:20:51+08:00 — Source used in the article above.
- 2019年SEO趨勢是什麼?「機器學習、物聯網、使用者體驗」影響下的SEOwww.awoo.ai · 2019-02-21T01:55:38+00:00 — Discusses: If no field data is displayed, does that mean the page has no problems?
- Google PageSpeed Insights: What It Is & How to Boost Your Scorewww.semrush.com · 2023-07-10T09:40:00Z — Source used in the article above.
- How To Audit Publisher Websites In 2026www.searchenginejournal.com · 2026-08-24T12:00:14+00:00 — Discusses: If no field data is displayed, does that mean the page has no problems?
