Page size breakdown
We download a page the way a first-time visitor’s browser would, with an empty cache, and count every byte that crosses the network. You get the total, where it comes from and what could go.
Every file has two sizes
There’s the size a file has on the network and the size it has once unpacked. Say your server compresses a 400 KB script down to 110 KB. The visitor pays 110 KB in download and mobile data, and their phone still has to read all 400 KB. This tool grades the first number, the transferred bytes, and shows the second next to it.
Bytes turn into waiting time fast. On the slow 4G profile Lighthouse uses for its tests, 1.6 megabits per second, a 2 MB page takes about ten seconds to download, and that’s before any script has run.
What gets downloaded
Once it has the HTML, the tool requests what a browser would request on a first load:
- stylesheets, and files announced with
rel="preload"ormodulepreload; - scripts with a
src, exceptnomodulefallbacks and script types a browser doesn’t run; - images: for
srcsetand<picture>it takes the largest candidate a desktop browser would pick, and it readsdata-srcon lazy-loaded images; - video posters, the favicon, and images in inline
styleattributes; - fonts declared in
@font-facerules (the WOFF2 file when there is one, twelve fonts at most) and stylesheets pulled in with@import.
A few limits keep the check short: 80 files, 6 MB per file, 15 MB in total. If a page goes past 80 files, the request count is shown as “≥”.
The grade scale
| Grade | Transferred size |
|---|---|
| A | up to 0.5 MB |
| B | up to 1 MB |
| C | up to 2 MB |
| D | up to 3.5 MB |
| E | up to 5 MB |
| F | more than 5 MB |
To give you an idea, the median mobile page measured by the HTTP Archive weighs about 2.3 MB, which gets a D here. So a page can be perfectly average for the web and still be slow on a phone.
What to do with the findings
- Images over 200 KB
- Nearly always an image served larger than it’s displayed. Export it at the displayed width, or twice that for high-density screens, and let
srcsetoffer smaller versions to phones. - JPEG or PNG images
- When the server allows it, the tool re-encodes the eight heaviest images as WebP (quality 80) and AVIF (quality 55) and shows the size of each result. That way the saving is measured on your own pictures. Serve the new format with a fallback:
<picture> <source srcset="hero.avif" type="image/avif"> <source srcset="hero.webp" type="image/webp"> <img src="hero.jpg" width="1200" height="630" alt="…"> </picture>
- Images without width and height
- This one isn’t about weight. It’s why text jumps around while the page loads. Add both attributes.
- Text files sent uncompressed
- HTML, CSS, JavaScript or SVG above 1.5 KB that arrived without gzip or Brotli. The compression and cache audit has the server settings.
- More than 1 MB of JavaScript, or 300 KB of fonts
- Take out the libraries and plugins nobody uses, load the rest only on the pages that need them, and keep two or three font weights in WOFF2.
The carbon figure, and what isn’t counted
The estimate follows version 4 of the Sustainable Web Design model. It counts 0.300 kWh for each gigabyte transferred, which covers data centers, networks and visitor devices along with their manufacture, and multiplies that by a world average of 494 g of CO₂e per kWh. That comes to about 0.15 g per megabyte. The comparison with a gasoline car uses 170 g per kilometer. Take it as an order of magnitude, good for comparing two versions of a page.
No JavaScript is executed. Whatever scripts load later, such as tag managers, ads, chat widgets and images added on scroll, is missing from the total. So are background images declared inside stylesheets and the content of embedded videos. The real weight is often higher than the figure you see.
Questions people ask
What is a good page size for a website?
Under 1 MB transferred is a comfortable target for a content page, and under 0.5 MB is excellent. Past 2 MB, visitors on mobile networks wait several seconds for the download alone.
How do I reduce the size of a web page?
Start with images, since they’re usually the biggest share: get the dimensions right, then switch to WebP or AVIF. After that, turn on compression for text files, remove unused scripts and cut down on font files. The “What to trim” block of the result lists these in order of impact for your page.
Why is the size different from the Network tab in Chrome DevTools?
DevTools counts everything a real browser loads, including files requested by scripts, at the size of your own window and often with a warm cache. This tool doesn’t run scripts and measures a first visit. Disable the cache in DevTools if you want a fair comparison.
Does page size affect SEO?
Not as a number by itself. But heavy pages tend to have a slow Largest Contentful Paint, and that’s one of the Core Web Vitals Google measures. You can check yours with the Core Web Vitals report.
Is AVIF better than WebP?
AVIF files are usually smaller at the same visual quality, especially for photographs, and they take longer to encode. Current versions of Chrome, Firefox and Safari read both. The conversion test shows which one wins for each of your images.
How accurate is the CO2 estimate?
It’s a model built on bytes transferred and global averages. Nobody measured your server or your visitors. It’s fine for comparing pages with each other and far too rough for carbon accounting.