On September 15, Google's Chrome team added four new experimental metrics to the Chrome User Experience Report (CrUX) — the same public dataset that powers Core Web Vitals performance rankings. The new metrics measure ad load from real Chrome user sessions across the open web.
The four metrics: Ad Count tracks the average number of ads visible in the viewport at one time. Ad Density measures what fraction of the visible screen area ads occupy. Ad Weight (Network) measures how many bytes ads consume per page load. Ad Weight (CPU) measures how many milliseconds of processor time ads take.
All four are reported at the 75th percentile across actual Chrome users. Coverage is limited to programmatic supply — sites with a valid ads.txt file listing authorized sellers. According to Search Engine Journal and Search Engine Land, there are no benchmarks and no thresholds established. These are measurements without a rating system.
Why "experimental" doesn't mean ignore it
Google's track record with CrUX data follows a clear pattern. The Core Web Vitals metrics — Largest Contentful Paint, Cumulative Layout Shift, Interaction to Next Paint — each started as informal Chrome experiments. Each became a confirmed ranking signal, and landing page quality in Google Ads now reflects that infrastructure.
None of that is a guarantee the ad load metrics follow the same path. But Google doesn't instrument Chrome at scale, build a BigQuery pipeline, and publish a developer API for data it has no intention of using. The question is when, not whether. And the structural logic is obvious: a page where ads dominate the viewport and consume significant processor time delivers a measurably worse user experience than one where they run lean. Google already cares about user experience in ad ranking. The gap between "we measure this" and "we use this" is smaller than it looks.
What the data actually gives you today
For anyone running PMax with Display enabled, or standalone Display Network campaigns, CrUX ad load data is the first independent publisher inventory audit that has existed at real scale.
Before September 15, you could pull a placement report and see the domains your ads ran on. You couldn't easily verify whether those domains were presenting users with two ads or twelve. Now there's a publicly available signal sourced directly from Chrome telemetry. A site in your placement report with high Ad Density and heavy Ad Weight (CPU) scores is a page where user attention is split across multiple competing ad calls and load performance suffers. You're paying the same CPM for worse conditions.
Publishers with cleaner inventory now have an independently sourced data point for direct-deal conversations. Low Ad Density relative to category peers is a measurable advantage they can now put a number on.
The PMax placement check worth running before Q4
PMax controls its own placement choices. You don't decide where Display impressions run — Google's optimization does. The placement report gives you the domain list. CrUX ad metrics give you a way to audit the quality of that list on dimensions Google doesn't surface in the campaign UI.
The practical version: pull your top 30 Display and PMax placements by spend from the last 90 days, look up their Ad Density and Ad Weight (CPU) scores in the CrUX API (BigQuery query structure is covered in the Chrome for Developers documentation), and compare placement-level conversion rate against those scores. If high-ad-density domains show lower conversion rates than your campaign average, you have a reason to exclude them. Exclusion lists work, placement-level bid adjustments work, and turning off Display in a PMax campaign entirely is a lever that some DTC accounts have used to stop the bleed in Q4.
This is not a fire drill. Nothing changed for active campaigns on September 15. But Q4 is three weeks away, and a placement quality audit is lower-stakes now than in October when every conversion counts. The CrUX data is faster and more objective than manually visiting fifty domains.
Google put publishers' ad loads on public record. That data is there to be used — and eventually, it will be.
Sources: Search Engine Land, Search Engine Journal, Search Engine Roundtable, Google, September 2026
