The 2025 Web Performance Benchmark: Mobile Still Has the Harder Job
A review of 2025 Core Web Vitals coverage across millions of websites, with practical implications for mobile performance work.
Data source: HTTP Archive and Chrome UX Report (CrUX), primarily July 2025
The mobile pass rate improved, yet half of measured sites still missed the full target
HTTP Archive’s 2025 Web Almanac reports that 48% of mobile websites and 56% of desktop websites achieved good performance across all Core Web Vitals. The mobile rate rose from 44% in 2024, while desktop moved from 55% to 56%. The direction is positive, but progress was much smaller on desktop and fewer than half of mobile sites met all three good thresholds in the dataset. The Performance chapter describes the analysis and metrics.
What the benchmark says
The report combines controlled page-load measurements from HTTP Archive with real-user experience data from CrUX. Its primary analysis uses a July 2025 crawl and covers millions of sites. The three measures span loading (LCP), interaction responsiveness (INP), and visual stability (CLS), so an overall pass requires a page to perform well across all three dimensions.
The gap between overall pass rates and LCP results is instructive. In the report, 62% of phone pages had a good LCP, compared with 74% on desktop. Since the overall Core Web Vitals result also includes responsiveness and layout stability, improving the first visible image alone cannot guarantee an overall pass.
What teams should do with the findings
Prioritize mobile pages and templates first when field data shows a mobile gap. Test on a constrained device and network profile, then inspect the request waterfall for late-discovered hero images, oversized media, and render-blocking assets. Measure interactions separately with field data; a page-load lab run cannot stand in for real interaction responsiveness.
Use the 75th percentile by device as the operational target, and track each metric independently. A single aggregate score can conceal a slow interaction or a layout issue even when loading is acceptable. After deploying a fix, check real-user data again and segment by template, geography, and device where sample size allows.
Google’s recommended good thresholds remain LCP ≤ 2.5 seconds, INP ≤ 200 milliseconds, and CLS ≤ 0.1 at the 75th percentile. Google’s documentation also cautions that Core Web Vitals alone do not guarantee a search ranking. Threshold methodology and page experience guidance provide the definitions and context.
Method and limitations
This is a synthesis of the published Web Almanac findings, not a Speed Test World measurement. The Web Almanac uses a defined crawl and Chrome UX Report data; results vary by device mix, site selection, and sample coverage. The percentages describe the sites and period analyzed, not every page view on the web or the performance of any specific site today. A site’s own field data and controlled tests are needed to diagnose an individual bottleneck.