Network resource types and the Core Web Vitals
Alongside HTML documents (99.9%), images (98.0%), scripts (95.2%), stylesheets (93.2%), and fonts (89.4%), 59.3% of audited sites (122,240) fire runtime fetch() requests (85% good LCP, 1.4s median) and 49.6% fire XMLHttpRequest calls (85% good LCP).
98.0% of sites use image requests network resource type mix, and 86% of those sites have a good LCP, against 86% of all sites.
LCP pass rate
Sites with image requests network resource type mix have a good LCP most often: 86%, against 86% of all sites. Sites with XHR requests resource types have it least often, at 85%.
Network resource type mix usage
HTML documents is the most used: 99.9% of sites. Image requests is 98.0% of sites.
LCP phase breakdown by network resource types loaded
Across the four LCP sub-parts in Chrome's field data, the largest gap between image requests and XHR requests sits in resource load delay: 274 ms median for image requests against 308 ms for XHR requests. Overall median field LCP is 1.4s for image requests (TTFB 573 ms, load delay 274 ms, load duration 110 ms, render delay 192 ms) versus 1.5s for XHR requests.
Understanding network resource types
How early fetch() and XHR API calls compete with critical rendering assets
Almost every web page loads an HTML document (99.9%), images (98.0%), scripts (95.2%), stylesheets (93.2%), and web fonts (89.4%, 86% good LCP at a 1.4s median). However, 59.3% of sites (122,240) also fire client-side fetch() requests during initial load (85% good LCP, 1.4s median) and 49.6% (102,368) fire XHR calls (85% good LCP, 1.5s median), often to fetch client-rendered content, A/B test configs, or cart state.
Render above-the-fold data server-side in the initial HTML document rather than waiting on client-side fetch() or XMLHttpRequest calls during startup, and defer telemetry beacons until after LCP. Across all audited sites, 98.0% load images (86% good LCP) while 59.3% fire fetch() calls during load (85% good LCP).
How network resource type mix plays out per platform
What the field data shows when you isolate a single CMS or framework.
WordPress 65,509 sites WordPress sites load scripts on 100.0%, images on 99.9%, CSS on 99.2%, and web fonts on 97.2%.
Across 65,509 WordPress sites, 100.0% load external JavaScript, 99.9% load images, 99.2% load CSS stylesheets, and 97.2% download web fonts (76% good LCP, 1.8s median).
Shopify 8,573 stores 99.9% of Shopify stores fire runtime fetch() calls alongside scripts (100.0%) and stylesheets (100.0%).
Across 8,573 Shopify storefronts, 99.9% execute fetch() requests during page load (97% good LCP, 1.3s median), 100.0% load scripts, and 100.0% load stylesheets.
React & Next.js 9,570 sites 91.4% of React & Next.js sites execute fetch() requests during load (80% good LCP, 1.5s median).
Across 9,570 React sites, 100.0% load JS bundles, 99.9% load images, 98.5% load web fonts (79% good LCP), and 91.4% fire fetch() requests (80% good LCP).
Wix 5,825 sites 100.0% of Wix sites use fetch() and beacon pings alongside scripts and images (1.2s median LCP).
Across all 5,825 Wix sites, 100.0% execute fetch() requests (92% good LCP, 1.2s median) and 100.0% send ping requests.
By Arjen Karel · 210,902 crawled sites (69,745 with desktop Chrome field data, 206,252 sites with network resource type mix, October 2026). How we measured.