Chrome adds four experimental ad metrics to its CrUX dataset

Google added four experimental advertising metrics to the Chrome User Experience Report on 15 September. Google published the change on its Chrome developer blog, in a post by Alex Cone. The four numbers describe how much of a page an ad load actually takes.

Ad count and ad density measure what fills the screen

Ad count is the average number of ads in the viewport, which is the part of a page a reader sees without scrolling. Ad density is the average share of that area covered by ads. Search Engine Land reports that Chrome samples the visible area once a second to work out both.

Two sites can carry the same number of ads and still score very differently. One large ad above the article gives a low count and a high density. That split is the reason Google publishes both numbers instead of one.

Ad weight arrives in bytes and in milliseconds

The other two metrics cover cost rather than space. Ad weight for network usage counts the data that ads pull down, in bytes. Ad weight for processor usage counts the time ads take to run, in milliseconds. Both build up across a session instead of being sampled, according to Search Engine Land.

Chrome decides which requests belong to ads by filtering network traffic and watching which scripts run. The figures therefore cover the whole ad setup on a page, not only the creative a reader sees.

Google sets no target for any of the four numbers

All four are labelled experimental. They sit outside Core Web Vitals, the speed measures Google uses in its own guidance, and Google has published no threshold for any of them. There is no pass mark to chase and no stated effect on ranking.

The data follows the usual rules of the report. The same dimensions apply, the same eligibility rules decide which sites appear, and results are reported at the 75th percentile. Google says the metrics are in the report now and that coverage should widen over the coming month.

Rivals can now look up each other’s ad load

The report is public, so a media owner can pull the same four numbers for a competitor without asking anyone. That changes the conversation in a sales meeting, and it also gives buyers a way to check what a site really loads before booking. Google made a similar move on counting when AdSense stopped counting ads that never render.

The most useful first check is the processor figure on a phone template, because that is where a heavy ad setup costs the most. Publishers already watching what AI answers do to their traffic, as Le Monde has been doing with its subscriber numbers, now have a second public gauge to sit beside it. Coverage is still filling in, so a site missing today may show up in October.