Subresource compression algorithms and the Core Web Vitals

87.6% of audited sites (181,998) serve Gzip-compressed subresources (86% good LCP), 66.8% serve Brotli (86% good LCP), 48.8% serve Zstandard (85% good LCP), and the 4.0% adopting Dictionary-Compressed Brotli (dcb) reach 97% good LCP (1.3s median).

Field data PhoneDesktopAll Scope All sites October 2026 edition · Desktop field outcomes

4.0% of sites use Dictionary-Compressed Brotli (dcb) subresource HTTP compression, and 97% of those sites have a good LCP, against 86% of all sites.

The State of Web Vitals · October 2026 · 212,308 crawled sites · 70,213 with desktop field data · 207,658 sites with subresource HTTP compressionBy Arjen Karel · corewebvitals.io/state-of-cwv
1

LCP pass rate

Sites with Dictionary-Compressed Brotli (dcb) subresource HTTP compression have a good LCP most often: 97%, against 86% of all sites. Sites with Zstandard (zstd) compression subresource compression have it least often, at 85%.

SitesGood, needs improvement, poor LCPMedian
All sites70,21386%1.4s
Dictionary-Compressed Brotli (dcb) 4.0%3,79297%1.3s
Unencoded binary assets (images / WOFF2) 99.6%69,91186%1.4s
Gzip (Content-Encoding: gzip) 87.6%62,24386%1.4s
Brotli (Content-Encoding: br) 66.8%49,73786%1.4s
Zstandard (Content-Encoding: zstd) 48.8%41,88385%1.4s
GoodNeeds improvementPoorSites: with Chrome field data
The State of Web Vitals · October 2026 · 212,308 crawled sites · 70,213 with desktop field data · 207,658 sites with subresource HTTP compressionBy Arjen Karel · corewebvitals.io/state-of-cwv
2

Subresource HTTP compression usage

Unencoded binary assets is the most used: 99.6% of sites. Dictionary-Compressed Brotli (dcb) is 4.0% of sites.

Sites using each option (sites can use more than one) 207,658 sites
Unencoded binary assets (images / WOFF2) 99.6% Gzip (Content-Encoding: gzip) 87.6% Brotli (Content-Encoding: br) 66.8% Zstandard (Content-Encoding: zstd) 48.8%
Unencoded binary assets (images / WOFF2) 99.6% Gzip (Content-Encoding: gzip) 87.6% Brotli (Content-Encoding: br) 66.8% Zstandard (Content-Encoding: zstd) 48.8% Dictionary-Compressed Brotli (dcb) 4.0%
The State of Web Vitals · October 2026 · 212,308 crawled sites · 70,213 with desktop field data · 207,658 sites with subresource HTTP compressionBy Arjen Karel · corewebvitals.io/state-of-cwv
3

LCP phase breakdown by subresource compression algorithm

Across the four LCP sub-parts in Chrome's field data, the largest gap between Dictionary-Compressed Brotli (dcb) and Zstandard (zstd) compression sits in resource load delay: 625 ms median for Dictionary-Compressed Brotli (dcb) against 290 ms for Zstandard (zstd) compression. Overall median field LCP is 1.3s for Dictionary-Compressed Brotli (dcb) (TTFB 306 ms, load delay 625 ms, load duration 152 ms, render delay 306 ms) versus 1.5s for Zstandard (zstd) compression.

Median LCP phase breakdown (Chrome field p75) Load delay Median LCP
Sites with subresource HTTP compression 51,310572 ms274 ms191 ms274 mslab 316 ms1.4s
Unencoded binary assets (images / WOFF2) 51,162572 ms275 ms191 ms275 mslab 316 ms1.4s
Gzip (Content-Encoding: gzip) 46,095580 ms280 ms196 ms280 mslab 323 ms1.4s
Zstandard (Content-Encoding: zstd) 31,989617 ms290 ms205 ms290 mslab 332 ms1.5s
Brotli (Content-Encoding: br) 37,243587 ms296 ms203 ms296 mslab 340 ms1.4s
Dictionary-Compressed Brotli (dcb) 3,101306 ms625 ms306 ms625 mslab 616 ms1.3s
TTFB Resource load delay Resource load time Element render delay
The State of Web Vitals · October 2026 · 212,308 crawled sites · 70,213 with desktop field data · 207,658 sites with subresource HTTP compressionBy Arjen Karel · corewebvitals.io/state-of-cwv
4

Understanding subresource compression algorithms

Brotli, Zstandard (zstd), and Dictionary-Compressed Brotli (dcb) across page subresources

Nearly every page (99.6%) transfers pre-compressed binary formats like WebP, AVIF, and WOFF2 without extra Content-Encoding. For text subresources (JS, CSS, SVG, JSON), 87.6% of sites use Gzip (86% good LCP, 1.4s median), 66.8% use Brotli (86% good LCP), 48.8% include Zstandard responses (85% good LCP, often from Meta/Google third-party scripts), and 4.0% (8,278 sites, led by Shopify/Cloudflare) use shared-dictionary Brotli (dcb) to achieve 97% good LCP (1.3s median).

Enable Brotli (br) and Zstandard (zstd) at your CDN edge for all text-based subresources (text/css, application/javascript, image/svg+xml, application/json) to shrink transfer size by 15% to 25% over Gzip. Across all audited sites, 66.8% serve Brotli subresources (86% good LCP) and 4.0% serve Dictionary-Compressed Brotli (97% good LCP).

The State of Web Vitals · October 2026 · 212,308 crawled sites · 70,213 with desktop field data · 207,658 sites with subresource HTTP compressionBy Arjen Karel · corewebvitals.io/state-of-cwv
Platform & Framework Takeaways

How subresource HTTP compression plays out per platform

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

Shopify 8,634 stores
100.0% / 95.5% Brotli / Dictionary-Compressed Brotli on Shopify

95.5% of Shopify stores use Dictionary-Compressed Brotli (dcb) alongside Brotli (100.0%), passing LCP at 97%.

Across Shopify storefronts, 100.0% serve Brotli (br) responses (97% good LCP) and 95.5% serve Compression Dictionary Transport (dcb) responses (97% good LCP, 1.3s median).

Read the Shopify subresource compression breakdown →
React & Next.js 9,658 sites
91.1% / 97.7% Brotli / Gzip subresources on React

91.1% of React & Next.js sites serve Brotli-compressed subresources (79% good LCP, 1.5s median).

Across 9,658 React sites, 91.1% serve Brotli-compressed bundles (79% good LCP), 97.7% use Gzip (79% good LCP), and 41.3% include Zstandard-compressed scripts (75% good LCP).

Related signals Scripts per page → Stylesheet origin (1P/3P) → DOM depth → Responsive image markup → By Arjen Karel · 212,308 crawled sites (70,213 with desktop Chrome field data, 207,658 sites with subresource HTTP compression, October 2026) · How we measured