PSI scores are useful for identifying opportunities to improve performance, but should not be treated as a search ranking report card. Not every page needs a perfect score. A 2023 Semrush article states that Google does not rank websites directly by their PSI scores. It suggests that improving the underlying loading speed and page experience may benefit search performance, but rankings also depend on content quality, backlinks, and technical SEO. This was the article’s explanation at the time. It should not be taken as a complete account of how rankings work today, let alone a guarantee that a higher score will produce higher rankings.
Scores are clues for improvement, not a seal of approval
The Semrush article cites these score ranges: 90 or above is good, 50 to 89 needs improvement, and below 50 is poor. These ranges can help readers understand the report, but crossing a score threshold should not be interpreted as gaining a ranking advantage.
The article also cautions that if many pages score below 90, this may indicate a poor user experience and may have a negative SEO impact. Neither outcome is inevitable. What visitors actually need is to find information or solve a problem quickly, not to see a website achieve a score of 100.
Distinguish test conditions from real-world usage
An iThome article distinguishes between two types of data: lab data is collected under preset device and network conditions and is useful for debugging, while real-user data reflects actual browsing conditions. Different devices, networks, and browsers produce different experiences, so a website loading quickly for you does not mean it does so for every visitor.
- Look at the problems as well as the overall score: Rather than focusing only on whether the score is high or low, identify specific issues such as content appearing slowly, delays in responding to interactions, or layout shifts.
- Make trade-offs based on the website’s purpose: A photography portfolio needs to preserve photo quality, so image optimization should account for presentation needs rather than simply aim to eliminate warnings.
This does not mean low scores can be ignored. It means deciding which recommendations to follow based on how the website is used, rather than applying them all without question.
Separate speed improvements from ranking promises
Text published on iThome makes stronger claims about scores and rankings: it describes a high score as indicating that a website’s architecture meets Google’s standards and says that scores affect keyword rankings. This differs from Semrush’s explicit distinction between “PSI scores” and “page speed.” The claims on iThome alone do not justify saying that Google ranks websites directly by their overall scores.
A practical starting point is to address specific user experience problems. For example, a 2022 article by David Design recommends reserving space for images and videos to reduce layout shifts as content loads. It also recommends displaying a default font while the intended font is still loading, so that text does not remain temporarily invisible. These recommendations aim to make the page more stable and reduce the wait for text to appear; they do not indicate how much rankings might improve.
When planning optimization work, first identify the problems you want to address. Set aside resources for content and other SEO work as well, rather than putting all your effort into chasing a perfect score.
PSI scores and their impact on search rankings: perspectives from across the web
What can LCP tell us?
LCP measures how long it takes for the largest content element on a page to render. That element may be an image, a video, or text. If the report indicates poor performance on this metric, first examine the page’s largest content element, then consider compressing images, streamlining CSS, or optimizing the critical rendering path. A 2022 article by David Design recommends keeping LCP below 2.5 seconds. This is the standard stated in that article; it cannot be used to infer ranking changes.
What is the difference between INP and FID?
Both relate to responsiveness to user input, but their scope differs: FID focuses on the delay following the first input, while INP looks at responsiveness throughout the user’s interactions. A 2022 article by David Design considers the latter more comprehensive and recommends reducing unnecessary code or third-party scripts. This can provide a starting point for investigating sluggish interactions, but the metrics alone cannot be used to predict rankings.
Do missing scores in bulk PSI data extraction mean a page performs poorly?
No. Screaming Frog’s 2018 article offers troubleshooting advice for the extraction workflow it describes. If a page renders normally but some scores are missing from the extracted data, first check the XPath and Extract Text settings. Then check PSI to see whether the data is available there. The article notes that much of the real-world data is available only for high-traffic pages. This advice applies to the workflow at that time; it does not describe current data availability rules.
References
- Google PageSpeed Insights: What It Is & How to Boost Your Scorewww.semrush.com · 2023-07-10T09:40:00Z — Source used in the article above.
- PageSpeed Insights (PSI) 是什麼? 對SEO有影響?www.da-vinci.com.tw · 2022-09-30T12:47:58+08:00 — Discusses: What can LCP tell us?; What is the difference between INP and FID?
- Bulk Testing PageSpeed Insights with the SEO Spiderwww.screamingfrog.co.uk · 2018-08-06T13:00:02+00:00 — Discusses: Do missing scores in bulk PSI data extraction mean a page performs poorly?
- 網站設計達成目標轉換的關鍵因素及技巧ithelp.ithome.com.tw — Source used in the article above.
- Lighthouse 和 PageSpeed,檢測網站該看哪個指標?ithelp.ithome.com.tw — Source used in the article above.




