What 615 Real Image Audits Found

Original research · Updated August 13, 2026 · Data collected 2025-02-27 to 2026-08-13

Between February 2025 and August 2026, 615 websites were run through our scanner, which loads a page in a real headless browser and measures every image twice: the pixel size of the file, and the size the browser actually drew it at. This is what those 16,110 images look like in aggregate.

How to quote these numbers

This is a field sample, not a random sample of the web. Every site here was submitted by someone who chose to audit it, and people audit pages they already suspect are heavy. That almost certainly pushes the oversized rate higher than the web at large.

Please frame any citation as "among 615 sites submitted to ImageDimensions' scanner" rather than "on the web". We would rather be quoted accurately than impressively.

Key findings

48.2%
of images are oversized

9,301 images had both a file size and a rendered size we could measure. 48.2% of them ship at least 4x more pixel area than they display — and that is after allowing a full 2x retina budget.

38.7%
have no alt text

4,990 of 12,898 HTML images carried a missing or empty alt attribute. Some of those are legitimately decorative; most are not.

3.45x
median overshoot

The typical measurable image carries about three and a half times the pixel area it needs. This is the median, so half of all images are worse than this.

63.7%
of sites have at least one

Of the 543 sites with measurable images, 63.7% ship at least one image that is 4x oversized. Oversized images are close to universal.

The tail is heavier than the median suggests: 33.2% of measurable images are oversized by 10x or more in area, and 61.7% exceed 2x.

38.7% of images have no alt text

This was not what we set out to measure, and it is the finding that surprised us most. Of 12,898 HTML images, 38.7% had a missing or empty alt attribute. An empty alt="" is valid and correct markup for a decorative image, so this figure is an upper bound on the problem rather than a count of defects. But decorative images are a minority of what appears on a typical page, and a rate this high across 615 sites suggests most of it is simply missing.

Format adoption is slower than you would think

By file extension, across all 16,110 images:

FormatImagesShare
PNG5,24832.6%
JPG4,46427.7%
SVG2,71316.8%
WebP2,03912.7%
extensionless / CDN8745.4%
GIF6914.3%
other430.3%
AVIF340.2%

Only 12.9% of images use a modern format (WebP plus AVIF), while PNG and JPG together still account for 60.3%. AVIF has barely registered at 0.2%. One important caveat: this counts file extensions, so it undercounts WebP and AVIF delivered by CDNs through content negotiation, where the URL keeps a .jpg extension. The 5.4% extensionless bucket hides more of the same.

One in five images is a CSS background

3,212 of 16,110 images (19.9%) were CSS background images rather than <img> elements. This matters methodologically: an audit that only parses HTML for <img> tags misses roughly a fifth of the images on a page, and CSS backgrounds skew toward large decorative hero art — exactly the heavy files an audit most needs to catch.

Median page: 5 images. Heaviest: 1,400.

The median page in the sample carried 5 images, but the mean was 26.2 and the largest single page held 1,400. That gap between median and mean is the whole story of image performance: most pages are modest, and a small number of extreme outliers drag every average upward. Quote the median.

Methodology

  • Collection: every URL submitted to the public scanner between 2025-02-27 and 2026-08-13.
  • Deduplication: one row per site, using only the most recent scan, so repeat audits of the same URL do not double-count.
  • Measurement: pages are loaded in a real headless browser, not parsed as static HTML, so JavaScript-inserted and lazy-loaded images are included. Each image is measured for its intrinsic pixel size and its rendered size.
  • "Oversized" definition: intrinsic pixel area divided by rendered pixel area. A 2x retina image is 4x area, so our 4x threshold does not penalise correct retina serving.
  • Measurable subset: 9,301 of 16,110 images had both dimensions available. The rest were hidden, failed to load, or were rendered at zero size.
  • Aggregate only: no individual site is named, and no per-site data is published.

Run the same audit on your site

Every figure above comes from the same free scanner. Point it at any URL and it reports intrinsic versus rendered size for every image on the page, CSS backgrounds included.

Scan a Page

Citation

Journalists and researchers are welcome to quote this freely. Suggested form:

ImageDimensions, "What 615 Real Image Audits Found: A Field Study"
(August 2026). Based on 16,110 images across 615 websites
submitted to the ImageDimensions scanner, Feb 2025 - Aug 2026.
https://imagedimensions.com/research/image-audit-field-study-2026

The underlying measurement is reproducible: the scanner that produced this data is free and public, and the CLI and MCP server run the same measurement from a terminal or an AI agent, so any figure here can be checked against your own sites.

Related