Chrome CrUX Ad Metrics: Measuring How Ad-Heavy Your Pages Really Feel

Chrome CrUX ad metrics are a new, experimental public dataset that measures how much of your page real users hand over to advertising. Google added four of them to the Chrome User Experience Report in September 2026: Ad Count, Ad Density, Ad Weight (Network Usage) and Ad Weight (CPU Usage) (Search Engine Journal). They are not part of Core Web Vitals, and they ship with no suggested targets. What they give you is a field number for something teams used to argue about from screenshots.

The dataset is public. That is the part front-end developers should notice first. You can read ad load for sites you do not own, including competitors, from the same source that feeds your Core Web Vitals dashboards.

What the four Chrome CrUX ad metrics measure

Each metric answers a different question about ad load.

• Ad Count is the average number of ads in the viewport.
• Ad Density is the average fraction of viewport area covered by ads.
• Ad Weight (Network Usage) is the bytes ads download, accumulated across the session.
• Ad Weight (CPU Usage) is the main-thread time ads consume, in milliseconds.

Two describe what the user sees. Two describe what the device pays for it. Density is the number most people read first, because it maps to the feeling of a cluttered page more directly than any waterfall does (Chrome for Developers).

How Chrome samples them, and why the 75th percentile matters

For Ad Count and Ad Density, Chrome inspects the visible viewport once per second and averages the samples. The two weight metrics accumulate over the whole session, so a long read on a heavy page keeps adding bytes and milliseconds until the user leaves. CrUX then reports every value at the 75th percentile, the same cut it applies to Core Web Vitals (Chrome for Developers).

That percentile choice carries the meaning. A reported Ad Density of 23 means 75% of qualifying sessions saw 23% or less of the viewport occupied by ads, while the worst quarter saw more (Search Engine Land). A median would have hidden that tail. The 75th percentile describes a bad-but-common session, which is precisely the visit you are at risk of losing. Read it the way you already read LCP.

Who shows up in the data

Not every site qualifies. Aggregate reporting is limited to sites carrying an ads.txt file that names at least one authorized seller (Search Engine Land). If a site runs third-party display ads and declares its sellers, it is in scope. If it does not, the ad metrics stay blank even where its Core Web Vitals are reported.

The metrics are live now in the CrUX API and the CrUX History API, with BigQuery support on the way for SQL queries across the full dataset (Chrome for Developers). History matters here. One reading is a snapshot. A trend line tells you whether a new ad unit quietly doubled your CPU cost last quarter.

Reading them next to Core Web Vitals

This is where a performance-minded team earns the data. The ad metrics do not replace Core Web Vitals. They explain them. Treat them as page performance signals that sit beside your vitals, not above them. When Interaction to Next Paint drifts upward and you cannot find the cause in your own JavaScript, Ad Weight (CPU) becomes a first-class suspect you can measure instead of guess. When layout feels crowded and engagement drops, Ad Density gives the complaint a number.

Read the pairing as a diagnosis, not a scorecard. High Ad Weight (CPU) against a poor INP points at ad scripts contending for the main thread. Rising Ad Density against falling scroll depth points at clutter over content. The discipline is the same one that separates a real cause from a coincidence in any technical SEO experiment: isolate the variable before you blame it.

One caution. These are new, experimental figures from a single upstream source, and Google can redefine or withdraw them. That is not hypothetical for anyone who watched GA4 and AdSense quietly change their own numbers this year (our breakdown). Version-stamp the date you pulled each reading, and never chart an experimental metric as though its definition is settled.

Where this leaves front-end teams

The honest summary is small and useful. For the first time, ad clutter has a real-user number that you and your stakeholders can see from the outside. It will not tell you what “good” looks like, because Google set no threshold. What it will do is settle arguments that used to run on opinion. If you own a page that carries ads, pull your four figures, note the date, and start a trend line before the next ad unit lands. That is the whole opportunity, and it is worth an afternoon.