About the Website Carbon Calculator
Enter a page and get the grams of CO2e one view produces, an A+ to F rating, a yearly total for your traffic, whether the host runs on verified green energy, and the specific files that account for most of it — each with the saving that fixing it would buy.
The arithmetic is the Sustainable Web Design Model, version 4, implemented exactly as the Green Web Foundation’s open-source co2.js library does it: energy per gigabyte for data centres, networks and devices, both operational and embodied, multiplied by grid carbon intensity. We checked our implementation against co2.js itself on dozens of byte sizes, green and grey, before publishing. The bytes are real: our server downloads the page and every file it references and counts what actually crossed the wire, compressed, headers included.
Two limits, stated up front. This is a static measurement — nothing runs JavaScript, so files a script fetches later are not counted and busy, script-heavy pages read lower than a browser would show. And the energy-per-byte figures are industry averages, which makes the result ideal for comparing pages and tracking your own progress, and only approximate as an absolute number.
The headline uses SWD v4’s global grid average (494 g CO2e per kWh). Electricity is far cleaner in France than in Poland, so where the server sits and where your visitors are changes the answer. Grid figures are Ember annual averages as shipped in co2.js.
A small badge for your footer that shows this page’s grams per view and rating, links to the full report, and re-measures itself every week so it never goes stale.
Reads the site’s sitemap and measures up to 25 of its pages one after another, so you see the average per view and which templates are the heavy ones. Allow a couple of minutes for 25 pages.
Every measurement you run is kept in this browser, so after you compress those images you can see exactly what changed for the same page. Results are normally reused for six hours; after a change, use “Re-measure now” to skip that. Nothing is uploaded.
How to Use the Website Carbon Calculator
Paste the address of a page and press Measure. Measure the pages people actually land on — a long article, a product page with a photo gallery, a checkout — rather than only the homepage, because that is where the weight usually is. Enter your monthly page views to turn grams per view into a yearly figure. Then go straight to Heaviest Files: it ranks every file by the emissions it adds and shows the saving from removing or halving it, which is the order in which to work.
The Sustainable Web Design Model, Version 4
The model, published by the Sustainable Web Design community and maintained in code by the Green Web Foundation, treats data transfer as the proxy for energy. Version 4 splits the system into three segments and counts two kinds of energy for each:
- Data centres — 0.055 kWh per GB to run, 0.012 kWh per GB embodied in the hardware.
- Networks — 0.059 kWh per GB to run, 0.013 kWh per GB embodied.
- User devices — 0.080 kWh per GB to run, 0.081 kWh per GB embodied.
Operational energy is multiplied by the grid intensity of each segment (494 g per kWh globally by default), embodied energy always by the global figure. Green hosting zeroes only the data centre’s operational share. A one-megabyte page therefore comes to 0.148 g on grey hosting and 0.121 g on green — exactly what co2.js returns for the same input. The rating bands are the model’s own: A+ up to 0.040 g, then A 0.079, B 0.145, C 0.209, D 0.278, E 0.359, and F above that.
Why Images Nearly Always Top the List
The HTTP Archive’s Web Almanac has found images to be the largest single category of page weight year after year, and much of it is avoidable: photos exported straight from a camera or a design tool, PNG used for photographs, and 2,400-pixel-wide images displayed in a 600-pixel column. Converting to WebP or AVIF and resizing to the displayed size usually shrinks such files dramatically with no visible difference — measure again afterwards and the history panel shows exactly how much. Our Image Compressor & Converter does both in your browser, and the heaviest-files table links to it next to every JPEG and PNG worth converting.
Images and embeds marked loading="lazy" are kept out of the headline, because a browser
fetches them only when a visitor scrolls near them — the headline is what every visit pays for on
arrival. They are still listed (tagged lazy), and the report gives a second figure for a visitor
who scrolls the whole page. Where a page uses <picture>, the file counted is the AVIF or
WebP source a current browser would choose, not the JPEG fallback.
Caching: the Return Visit Is a Different Page
A returning visitor’s browser reuses anything it is allowed to keep. The return-visit figure reads
each file’s Cache-Control, Expires, ETag and
Last-Modified headers and counts what would cross the wire again the next day: nothing for
files cached for a day or more, a few hundred bytes for files that can be revalidated, the full file for
everything else. A large gap between the first and return figures means caching is doing its job; a small
gap usually means static files are being served with no cache lifetime at all, which is a one-line server
fix.
What This Measurement Cannot See
The scan reads the HTML and the stylesheets the server returns. It does not run JavaScript, so files a script requests after the page loads — advertising, chat widgets, analytics beacons, anything a tag manager injects — are not counted. For a lean, mostly static page that changes little; for an ad-supported page it can miss a large share of the weight. Video files are not counted unless a page preloads them. If the site you are measuring blocks automated requests, the result reflects whatever page it sent us instead, and the report flags it.
For the bigger performance picture, run the Core Web Vitals Checker. To see what a site is built with, the Website Tech Stack Detector names it; to see who a page reports its visitors to, the Website Privacy X-Ray lists every third party. Browse every Dev & Tech tool for more.
Frequently Asked Questions
How can a web page produce carbon dioxide at all?
Every byte you serve is stored and sent by a data centre, carried across networks and processed by the visitor's device, and all three draw electricity. The Sustainable Web Design Model puts numbers on that chain: about 0.194 kWh of operational energy and 0.106 kWh of embodied energy (the share of building the hardware) per gigabyte transferred. Multiply by how much CO2 the electricity grid emits per kWh and you have grams per view.
Why is the figure an estimate rather than a measurement?
The bytes are measured; the energy per byte is not. Nobody can meter the share of a router or a phone that one page view uses, so the model uses industry-wide averages per gigabyte and an average grid intensity. That makes it excellent for comparing pages and tracking your own improvements, and unreliable as a precise absolute number. Treat a result of 0.20 g as "about a fifth of a gram", not as three significant figures.
Does green hosting make my site carbon neutral?
No. In the SWD v4 model green hosting removes only the data centre's operational emissions, which is roughly 18% of the total for a page. The network, the visitor's device and the embodied emissions of all the hardware remain. Green hosting is worth having, but shrinking the page reduces every segment at once, which is why page weight is the lever that matters most.
Why does my result differ from other website carbon calculators?
Three reasons. Some still use the older SWD v3 model, whose constants and rating bands are different. Some load the page in a real browser and count resources that JavaScript fetches later, which this static measurement cannot see. And some apply caching assumptions to the headline figure. Here the headline follows SWD v4's default — every view treated as a first visit — and the return-visit figure is shown separately, worked out from each file's actual caching headers.
What should I fix first to cut my page's emissions?
Whatever the heaviest-resources table puts at the top, which is nearly always images. Serving JPEG and PNG photos as WebP or AVIF at the size they are displayed is usually the single biggest saving available. After that: drop unused JavaScript, self-host and subset web fonts, make sure text files are compressed, and give static files a long cache lifetime so return visitors download almost nothing.