HTML compression (Brotli, Gzip, Zstandard) and the Core Web Vitals

Brotli (br), Gzip, Zstandard (zstd), or uncompressed HTML: how document compression affects Time to First Byte (561 ms on Brotli vs 706 ms on Gzip) and Largest Contentful Paint.

Field data PhoneDesktopAll Scope All sites November 2026 edition · Phone field outcomes

One in five sites, 20.6%, have no HTML compression. Of the 79.4% that do, 36% use Brotli, and 86% of those sites have a good LCP, against 82% of all sites.

The State of Web Vitals · November 2026 · 225,631 crawled sites · 135,091 with phone field data · 220,867 sites with HTML compressionBy Arjen Karel · corewebvitals.io/state-of-cwv
1

LCP pass rate

Sites with Brotli HTML compression have a good LCP most often: 86%, against 82% of all sites. Sites with Zstandard HTML compression have it least often, at 78%.

SitesGood, needs improvement, poor LCPMedian
All sites135,09182%1.6s
Brotli (br) 28.3%38,75586%1.5s
Gzip 45.9%62,08281%1.6s
Uncompressed (none) 20.6%27,10780%1.6s
Zstandard (zstd) 5.1%7,07178%1.7s
GoodNeeds improvementPoorSites: with Chrome field data
The State of Web Vitals · November 2026 · 225,631 crawled sites · 135,091 with phone field data · 220,867 sites with HTML compressionBy Arjen Karel · corewebvitals.io/state-of-cwv
2

HTML compression usage

Four in five sites uses HTML compression, and three in five of those use Gzip. Brotli is 36% of the sites with HTML compression, 28.3% of all sites.

Sites with HTML compression 220,867 sites
Uncompressed (none) 20.6% HTML compression 79.4%
Sites with HTML compression 175,322 sites
Gzip 57.8% Brotli (br) 35.7%
Gzip 57.8% Brotli (br) 35.7% Zstandard (zstd) 6.5%
The State of Web Vitals · November 2026 · 225,631 crawled sites · 135,091 with phone field data · 220,867 sites with HTML compressionBy Arjen Karel · corewebvitals.io/state-of-cwv
3

Where HTML compression shows up in the four LCP phases

Across the four LCP sub-parts in Chrome's field data, the largest gap between Brotli and no sits in time to first byte (TTFB): 561 ms median for Brotli against 644 ms for no. Overall median field LCP is 1.5s for Brotli (TTFB 561 ms, load delay 396 ms, load duration 165 ms, render delay 239 ms) versus 1.6s for no.

Median LCP phase breakdown (Chrome field p75) Load delay Median LCP
Sites with HTML compression 98,583649 ms336 ms236 ms336 mslab 326 ms1.6s
Gzip 45,043706 ms296 ms240 ms296 mslab 288 ms1.6s
Zstandard (zstd) 5,293758 ms308 ms254 ms308 mslab 298 ms1.7s
Uncompressed (none) 18,922644 ms353 ms217 ms353 mslab 338 ms1.6s
Brotli (br) 29,273561 ms396 ms165 ms239 ms396 mslab 392 ms1.5s
TTFB Resource load delay Resource load time Element render delay
The State of Web Vitals · November 2026 · 225,631 crawled sites · 135,091 with phone field data · 220,867 sites with HTML compressionBy Arjen Karel · corewebvitals.io/state-of-cwv
4

Understanding HTML document compression

Why Brotli HTML compression speeds up resource discovery

Brotli uses a built-in dictionary of common HTML, CSS, and JavaScript tokens, shrinking HTML documents 15 to 20 percent more than Gzip at comparable CPU cost. Because the browser's preload scanner parses the HTML stream chunk by chunk as it arrives, a smaller compressed document delivers the <head> and hero <img> markup in fewer TCP or QUIC packets. Sites serving Brotli HTML record a 561 ms median TTFB and 1.5s median LCP, compared with 706 ms TTFB and 1.6s LCP on Gzip.

Enable Brotli (br) compression on your HTML responses and keep Gzip as a fallback. Across all audited sites, 28.3% serve Brotli-compressed HTML and reach 86% good LCP (1.5s median), beating Gzip (45.9% of sites, 81% good LCP) and uncompressed HTML (20.6% of sites, 80% good LCP). For dynamic HTML generated on the fly, Brotli quality level 4 or 5 compresses faster AND smaller than Gzip level 6 without adding server CPU delay.

The State of Web Vitals · November 2026 · 225,631 crawled sites · 135,091 with phone field data · 220,867 sites with HTML compressionBy Arjen Karel · corewebvitals.io/state-of-cwv
Platform & Framework Takeaways

How HTML compression plays out per platform

What the field data shows when you isolate a single CMS or framework.

WordPress 70,035 sites
78% vs 66% good LCP (Brotli vs gzip)

50.1% of WordPress sites compress HTML with gzip (66% good LCP vs 78% with Brotli, 12 points apart).

Across WordPress sites, 50.1% serve gzip-compressed HTML (66% good LCP, 2.0s median), 31.5% serve Brotli-compressed HTML (78% good LCP, 1.8s median), and 12.3% send uncompressed HTML (64% good LCP).

Read the WordPress HTML compression breakdown →
React & Next.js 10,308 sites
91% vs 67% good LCP (Brotli vs gzip)

React & Next.js sites on Brotli (71.0%) pass LCP 24 percentage points more often than sites on gzip (17.8%).

Across React sites, 71.0% serve Brotli-compressed HTML (91% good LCP, 1.2s median) compared with 17.8% serving gzip-compressed HTML (67% good LCP, 2.0s median).

Related signals HTTP protocol mix → Connection hints → Cache strategy mix → Uncompressed text responses → By Arjen Karel · 225,631 crawled sites (135,091 with phone Chrome field data, 220,867 sites with HTML compression, November 2026) · How we measured