CSS hints

CSS performance hints in use: content-visibility, contain, will-change.

Field data PhoneDesktopAll Scope All sites Q1 2026 edition · All devices field outcomes
Metric LCP INP CLS
1

At a glance the headline numbers for CSS hints

CSS performance hints in use: content-visibility, contain, will-change.

5
Categories
In the distribution
32.5%
Fleet share
Top: will_change
49.3%
Sites with any
Of will_change

12.4% of sites use CSS containment. content-visibility sits at 11.2%.

The State of Web Vitals · Q1 2026 · 94,910 sites · all devices field datacorewebvitals.io/state-of-cwv
2

The CSS hints mix who uses what, and how fast each group loads

Median LCP (sites using feature)
0
0.5s
1s
1.5s
2s
2.5s
3s
3.5s
4s
Will change1.7s33% of sites
Aspect ratio1.7s29% of sites
Scroll driven1.6s15% of sites
Containment1.7s12% of sites
Content visibility1.8s11% of sites
VariantShare of sitesMedian
Will change
33%
1.7s
Aspect ratio
29%
1.7s
Scroll driven
15%
1.6s
Containment
12%
1.7s
Content visibility
11%
1.8s

CSS hints. On the fleet: 32.5% will change, 29.4% aspect ratio, 14.5% scroll driven. 49.3% of sites use at least one will_change.

The State of Web Vitals · Q1 2026 · 94,910 sites · all devices field datacorewebvitals.io/state-of-cwv
3

Why this matters for the Core Web Vitals, and where to start fixing it

These CSS properties limit how far a change can spread. contain declares a subtree self-contained, so a change inside it does not force layout outside it. content-visibility goes further and skips rendering offscreen sections until the visitor approaches them. will-change moves an animated element onto its own layer, so animating it does not repaint the rest of the page.

They attack the DOM problem directly: a big tree hurts because every change can touch all of it, and containment makes updates local. That is exactly what INP needs. will-change is the one to ration. Every promoted layer costs memory, and a blanket will-change makes performance worse, not better.

How does this affect the Core Web Vitals?

The choice barely moves the INP: 96% pass at best, 95% at worst. This signal does not separate passing sites from failing ones.

Related signals Scheduling API use → Interaction invoker types → Yielding strategy → INP phase breakdown → Chrome field data from 94,910 sites, representing millions of real page loads · How we measured