65.7% of sites declare no speculation rules at all

1

At a glance the headline numbers for Speculation posture

The strongest speculation rule the document declares: none, prefetch, prerender-until-script, prerender.

4

Speculation posture who uses what, and how fast each group loads

Median LCP

0

0.5s

1s

1.5s

2s

2.5s

3s

3.5s

4s

VariantShare of sitesMedian

prerender_until_script 491 sites

0%

1.2s

none 121,738 sites

66%

1.3s

prerender 908 sites

1%

1.5s

prefetch 62,089 sites

34%

1.7s

Prerender_until_script sites post the best median LCP (1.2s). Prefetch sites trail at 1.7s. Correlation, not causation. computed

The State of Web Vitals · Q2 2026 · 189,915 sites · desktop field datacorewebvitals.io/state-of-cwv

3

Passing LCP per speculation posture which group passes the LCP most often

VariantSitesPassing LCPLCP

Prerender until script 0.3% 491 93% 1.2s

None 65.7% 121,738 89% 1.3s

Prerender 0.5% 908 86% 1.5s

Prefetch 33.5% 62,089 78% 1.7s

Good Needs Improvement Poor Sorted best-passing first · median colored by its own rating · pass = good LCP (2.5s at p75) · one value per site

93% of Prerender until script sites pass LCP. Prefetch trails 15 points behind. computed

The State of Web Vitals · Q2 2026 · 189,915 sites · desktop field datacorewebvitals.io/state-of-cwv

4

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

Speculation rules let the browser start the next navigation before the click. Prefetch downloads the next document ahead of time. Prerender goes further and builds the whole next page in a hidden renderer, so navigating to it is a swap instead of a load. Prerender-until-script is the cautious middle: build the page, hold the scripts.

Done right, the next page's LCP is near-instant because it painted before the visitor arrived. The cost is speculative work: prerendering pages nobody visits burns server and client resources. That trade is what eagerness controls.

How does this affect the Core Web Vitals?

Speculation posture correlates with the LCP. With Prerender until script, 93% of sites pass the LCP. With Prefetch, 78% do. The causal link is weak: the choice mostly marks what kind of site made it.

The split is bigger on CLS. With Prerender until script, 97% of sites pass it. With None, 78% do.

Go fix it Speculation rules → The prerender-until-script speculation rule → View transitions and web performance →