At a glance the headline numbers for Media per page
Audio and video elements on the page and their weight.
The typical page embeds 0 media files.
Distribution & median INP site count and median INP at each level of media per page — n
Passing INP by media per page — n which level passes the INP most often
Media per page — n 0. p75 0. p99 6. Spearman with INP r = +0.40. At the low end (0): INP 48ms. At the high end (>p98): INP 65ms. computed
Distribution & median INP site count and median INP at each level of media per page — size
Passing INP by media per page — size which level passes the INP most often
Media per page — size 39 KB. p75 2.8 MB. p99 44.3 MB. At the low end (0 KB): INP 53ms. At the high end (>p98): INP 61ms. computed
Why this matters for the Core Web Vitals, and where to start fixing it
When audio or video is present, it outweighs everything else on the page. An autoplaying background video competes with the LCP image for bandwidth at the worst moment, and on phones it burns data the visitor never asked for.
A poster image carries the experience until the visitor opts in. It loads like an image, it can be the LCP candidate, and the video starts on interaction. Autoplay hero video is the expensive way to do what a good photo does.
How does this affect the Core Web Vitals?
Media per page correlates with the LCP. Media count separates passing sites from failing sites more than media weight does. Where the media count is low, 86% of sites pass the LCP. Where it is high, 81% do. The decline is gradual. There is no point where sites suddenly start failing.
Chrome field data from 189,915 sites, representing millions of real page loads. How we measured.