The State of Web Vitals

Bytes by resource type

The page weight split by resource type: HTML, JS, CSS, images, fonts, media.

Field data PhoneDesktopAll Scope All sites Q2 2026 edition · Desktop field outcomes
Metric LCP INP CLS
1

At a glance the headline numbers for Bytes by resource type

The page weight split by resource type: HTML, JS, CSS, images, fonts, media.

13
Categories
In the distribution
31.7%
Fleet share
Top: script
—
Sites with any
Of script

Scripts are 31.7% of all requests. Images are 30.5%.

The State of Web Vitals · Q2 2026 · 189,915 sites · desktop field datacorewebvitals.io/state-of-cwv
2

The bytes by resource type mix who uses what, and how fast each group loads

VariantShare of requestsMedian
Script
32%
Image
31%
Stylesheet
13%
Other
6%
Font
5%
Fetch
5%
Xhr
2%
Document
2%
Ping
2%
Preflight
1%
Media
0%
Manifest
0%
Prefetch
0%

Bytes by resource type. On the fleet: 31.7% script, 30.5% image, 13.4% stylesheet.

By count script leads (31.7%); by bytes it is image (63.3%). computed

The State of Web Vitals · Q2 2026 · 189,915 sites · desktop field datacorewebvitals.io/state-of-cwv
3

Passing INP per bucket every category and count level at once - color is the pass rate

1
2
3
4
5
6
7
8
9
10
11
12
Script 31.7%
99
99
99
99
99
99
99
99
99
99
99
98
Image 30.5%
98
99
99
99
99
99
99
99
99
99
99
98
Stylesheet 13.4%
99
99
99
99
99
99
99
99
99
99
99
99
Other 6.2%
99
98
98
98
93
89
100
96
94
96
100
99
Font 5.3%
98
99
99
99
99
99
99
99
98
99
99
97
Fetch 5.1%
99
99
99
99
98
99
99
98
97
99
97
96
Xhr 2.5%
99
99
99
99
99
99
97
97
98
98
98
96
Document 2.1%
99
99
99
99
99
Ping 1.6%
99
99
98
98
99
99
99
99
98
98
99
98
Preflight 1.1%
98
98
98
98
98
98
99
99
98
96
98
97
Media 0.3%
99
99
99
98
100
100
99
100
100
99
Manifest 0.1%
98
99
Prefetch 0%
100
100
← few of this category on the pagemany →
60%95%+ of sites passing INP Faded cells: under 100 sites

Each row is a category, each column its own count bucket (few on the left, many on the right); the cell is the share of those sites passing INP.

No category moves the INP pass rate much, however many a site ships. computed

The State of Web Vitals · Q2 2026 · 189,915 sites · desktop field datacorewebvitals.io/state-of-cwv
4

Few vs many - does quantity cost INP? the pass rate with few vs many of each category

60%70%80%90%100% few → many
Fetch 5.1% 99% → 96%
Xhr 2.5% 99% → 96%
Preflight 1.1% 98% → 97%
Script 31.7% 99% → 98%
Font 5.3% 98% → 97%
Ping 1.6% 99% → 98%
Other 6.2% 99% → 99%
Document 2.1% 99% → 99%
Media 0.3% 99% → 99%
Prefetch 0% 100% → 100%
Image 30.5% 98% → 98%
Stylesheet 13.4% 99% → 99%
Manifest 0.1% 98% → 99%
% of sites passing INP · hollow ring = pages with few, solid dot = pages with many

Per category: the pass rate among pages with FEW of it (hollow ring) against pages with MANY (solid dot), worst trend first. Thin buckets are excluded from the endpoints.

More Fetch costs the most: the INP pass rate falls from 99% with few to 96% with many. computed

The State of Web Vitals · Q2 2026 · 189,915 sites · desktop field datacorewebvitals.io/state-of-cwv
5

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

The total page weight tells you the size of the problem. The split per resource type tells you which Core Web Vital pays for it. Image bytes mostly cost LCP. They compete for bandwidth with the hero image and make it load slower. Script bytes are the expensive ones. After the download the main thread still has to parse and execute them, so a script-heavy mix shows up in INP. CSS blocks rendering until it is loaded. Font bytes delay text rendering.

That is why the same total weight can produce very different vitals. A content site with mostly image bytes needs modern formats and lazy loading. A web app with mostly script bytes needs deferral and dead code removal. The mix tells you which problem you have.

Go fix it Optimize images for Core Web Vitals → 14 methods to defer JavaScript → Remove unused CSS →
Related signals Cookies per site → Media per page → Responsive image markup → font-display strategy → Chrome field data from 189,915 sites, representing millions of real page loads · How we measured