Image compression is a practical web-performance issue because images appear on nearly every page and can vary dramatically in dimensions, format, and file size. In June 2024, the HTTP Archive found that 99.9% of more than 10 million scanned and parsed pages requested at least one image resource. The same measurement shows a median of 13 mobile images per page, a median mobile image size of 12 KB, and a median density of 2.1 bits per pixel.
Contents
- How common images are on the web
- How many images pages contain
- Pixel counts and image dimensions
- Format shares and adoption
- File-size distributions
- Compression density by format
- Animated GIF statistics
How common images are on the web
The first image compression statistic is about reach: images are not a niche page element. According to Media | 2024 | The Web Almanac by HTTP Archive, 99.9% of more than 10 million scanned and parsed pages requested at least one image resource in June 2024. That result describes pages observed by the crawl, not every page on the internet, but it establishes how broadly image delivery matters in the measured web.
The measurement period is important. These figures describe June 2024, while the reference-model figures later in this article describe 2024. They should be read as a snapshot of web practice during that period rather than as a timeless limit on image sizes or format usage.
Image-heavy pages were common even before considering how many pixels or bytes each image contained. At the 10th percentile, a page had one image element on both mobile and desktop. At the 25th percentile, pages had five image elements on mobile and six on desktop. The median was 13 on mobile and 14 on desktop. At the 75th percentile, the counts rose to 29 on mobile and 32 on desktop. At the 90th percentile, the counts were 56 on mobile and 62 on desktop. These are element counts, so they describe how many image elements were found, not how much image data a page transferred.
How many images pages contain
The percentile pattern shows a wide spread between lightly illustrated pages and highly visual pages. A page at the mobile 10th percentile had one image element, while a page at the mobile 90th percentile had 56. The maximum reported mobile-page count was 2,174 image elements in June 2024. A single extreme page should not be treated as typical, but it demonstrates why an average alone would hide the range of web image usage.
The desktop counts were slightly higher at each reported upper percentile: 14 versus 13 at the median, 32 versus 29 at the 75th percentile, and 62 versus 56 at the 90th percentile. The data does not say that desktop images were always larger or heavier; it reports the number of image elements found in the respective crawls. For compression work, page image count and per-image optimization are separate questions.
The table below keeps the mobile and desktop counts together for the three central comparison points.
| Page percentile | Mobile image elements | Desktop image elements |
|---|---|---|
| 25th | 5 | 6 |
| 50th, or median | 13 | 14 |
| 75th | 29 | 32 |
| 90th | 56 | 62 |
All values in this table come from the June 2024 HTTP Archive measurement reported by The Web Almanac. The mobile crawler’s maximum of 2,174 image elements is a separate extreme observation, not a percentile.
Pixel counts and image dimensions
Image dimensions help explain why two files with the same format can have very different compression outcomes. In June 2024, the 10th-percentile mobile image contained 0.001 megapixels. At the 25th percentile, the value was 0.013 megapixels. The median was 0.058 megapixels, followed by 0.262 megapixels at the 75th percentile and 0.778 megapixels at the 90th percentile.
Those values describe the number of pixels in a mobile web image, expressed in megapixels. They are not claims about a particular width-and-height combination, and they should not be converted into a specific display size without additional information about aspect ratio and rendering.
The largest image on each mobile page followed a different distribution. At the 25th percentile, the largest image measured 0.138 megapixels. At the median, it measured 0.540 megapixels. At the 90th percentile, it measured 3.130 megapixels. This page-level maximum is useful when looking for oversized outliers: the typical image and the largest image on the same page represent different optimization targets.
The gap between the median image size, 0.058 megapixels, and the median page-level largest image, 0.540 megapixels, also shows why a page audit should inspect individual assets rather than assume that every image follows the median. The figures come from the same June 2024 HTTP Archive mobile measurement and are not a forecast of future camera or screen resolutions.
Format shares and adoption
JPEG and PNG remained the two largest named formats in the June 2024 mobile crawl. JPEG represented 32.4% of mobile web images, while PNG represented 28.4%. GIF accounted for 16.8%, WebP for 12.0%, and SVG for 6.4%. ICO represented 1.3%, and AVIF represented 1.0%.
| Mobile image format | Share in June 2024 |
|---|---|
| JPEG | 32.4% |
| PNG | 28.4% |
| GIF | 16.8% |
| WebP | 12.0% |
| SVG | 6.4% |
| ICO | 1.3% |
| AVIF | 1.0% |
The shares are format observations from the mobile crawl, not quality rankings. A format’s share does not by itself show whether its files were smaller, sharper, or more suitable for a particular image. It only shows how often that format appeared in the measured set.
AVIF usage was 386% higher in 2024 than in 2022 in the mobile crawl. This is an adoption change across the reported period, not a statement that AVIF was the most common format in June 2024. AVIF’s measured 2024 share was still 1.0%, while JPEG, PNG, GIF, WebP, and SVG each had larger reported shares.
File-size distributions
The mobile image file-size distribution also had a long spread. At the 10th percentile, a mobile image weighed 0 KB in June 2024. The 25th-percentile value was 2 KB, the median was 12 KB, the 75th-percentile value was 56 KB, and the 90th-percentile value was 177 KB. The reported 0 KB value should be treated as the crawl’s displayed measurement at that percentile; it does not establish that every such image had literally no bytes.
The largest image per mobile page was much heavier at the upper end. Its 10th-percentile value was 6 KB, its 25th-percentile value was 37 KB, and its median was 135 KB. At the 75th percentile, the largest image weighed 404 KB, and at the 90th percentile it weighed 1,002 KB.
These two distributions answer different questions. The per-image distribution describes a randomly positioned mobile image in the crawl. The page-level largest-image distribution describes the heaviest image found on each page. A site can therefore have many small decorative or interface images while still carrying a large hero or product image that dominates the page’s image payload.
For photo editing workflows, the distinction is useful when choosing what to inspect first. A median file of 12 KB does not make a 1,002 KB 90th-percentile page maximum unimportant. Conversely, the maximum does not describe the typical asset. Both values are needed to see the range.
Compression density by format
Bits per pixel provide a way to compare file weight with pixel count. The 2024 reference model characterizes an uncompressed 8-bit RGB bitmap as 1:1 compression and 24 bits per pixel. Losslessly compressed RGB is characterized by an approximately 2:1 ratio and 12 bits per pixel. A 1990s-era lossy RGB rule of thumb is characterized by an approximately 10:1 ratio and 2.4 bits per pixel.
The observed mobile web distribution was broader than those simple reference points. Mobile web images measured 0.1 bits per pixel at the 10th percentile, 0.9 at the 25th percentile, 2.1 at the median, 5.6 at the 75th percentile, and 12.9 at the 90th percentile in June 2024. These are measured web-image densities, not fixed properties of a file format.
Format-specific median values add detail:
| Format | Median mobile bits per pixel |
|---|---|
| WebP | 1.3 |
| AVIF | 1.4 |
| JPEG | 2.0 |
| PNG | 3.8 |
| GIF | 6.7 |
The median values show that the measured WebP and AVIF images had the lowest listed densities in this comparison, at 1.3 and 1.4 bits per pixel. JPEG measured 2.0, PNG 3.8, and GIF 6.7. Lower bits per pixel indicate less file data per pixel in this measurement, but they do not independently describe visual quality. Image content, dimensions, encoder settings, animation, transparency, and acceptable artifacts all affect the result.
Animated GIF statistics
GIF deserves separate treatment because animation changes the amount of image data represented by a file. In June 2024, 32% of mobile GIFs were animated. Among animated mobile GIFs, bits per pixel measured 4.5 at the 10th percentile, 12.1 at the 25th percentile, 32.6 at the median, 60.6 at the 75th percentile, and 82.6 at the 90th percentile.
Non-animated mobile GIFs were lower at every reported percentile: 0.9 bits per pixel at the 10th percentile, 1.8 at the 25th percentile, 3.5 at the median, 5.7 at the 75th percentile, and 9.3 at the 90th percentile.
| Mobile GIF type | 10th percentile | Median | 90th percentile |
|---|---|---|---|
| Animated | 4.5 bits per pixel | 32.6 bits per pixel | 82.6 bits per pixel |
| Non-animated | 0.9 bits per pixel | 3.5 bits per pixel | 9.3 bits per pixel |
The animated and non-animated GIF measurements are not interchangeable. In the HTTP Archive figures, the median animated GIF measured 32.6 bits per pixel versus 3.5 for a non-animated GIF. The comparison describes compression density in the crawl; it does not assign a universal quality threshold or prove that animation is the only reason for every difference.
Taken together, the 2024 statistics describe a web where images are nearly universal, page image counts vary sharply, and file density depends heavily on the asset and format. The practical measurement questions are therefore specific: how many images a page requests, how many pixels each contains, how large the files are, and whether the selected format is carrying still or animated content.