What a Website Carbon Test Measures, and What It Cannot Tell You

Type a web address into a carbon checker and a few seconds later you get a figure such as “0.15 g CO₂e per visit” and a letter grade. It is tempting to read that as a measurement, like a reading from an electricity meter. It isn't. Nobody can see the electricity used by the servers, networks and phones involved in one page view. What a carbon test can do is measure something real about your page, how much data it sends, and run that through a published model. This article explains what goes in, what comes out, and where the limits are, so you can use the number with confidence instead of over-reading it.

What is actually measured

The one thing GreenTracer measures is page weight: the number of bytes transferred when the page loads. We ask Google PageSpeed Insights to load your page twice, once as a mobile visitor and once as a desktop visitor. Each run is a Lighthouse test, and Lighthouse reports a figure called total byte weight: the size of everything the page requested, including images, scripts, fonts, video and requests to other companies' servers such as analytics, chat widgets and embedded maps.

A few details matter when you read the result:

• It is a first, uncached visit. The browser starts with nothing saved, so every file is downloaded. A returning visitor usually downloads less because their browser already has some files.
• It is the initial load only. The test doesn't scroll or click, so images set to load lazily further down the page, and anything that appears after a click, are usually not counted.
• Bytes are counted as they travel. The figure is the compressed size sent over the network, not the size of the files on your server.
• Third parties count. If your page loads a video player or tracking script from someone else, those bytes are part of your page's weight.
• We use the heavier run. Mobile and desktop versions often differ, for example because desktop gets larger images. GreenTracer uses the heavier of the two. That is our own cautious choice, not a requirement of the model.

If the page can't be loaded, because it is private, blocked or times out, GreenTracer shows an error and saves nothing. We never fill the gap with a guessed page weight.

How bytes become grams

To turn page weight into emissions, GreenTracer uses the Sustainable Web Design Model, version 4 (SWDM v4). It estimates that moving one gigabyte of data uses 0.300 kWh of electricity, shared across three parts of the system. Each part has an operational share (electricity used while running) and an embodied share (energy used to make the equipment):

That energy is multiplied by a global average grid intensity of 494 g CO₂e per kWh, the default in SWDM v4. Gigabytes are decimal, so 1 MB is 0.001 GB. For a page that transfers 1 MB:

0.001 GB × 0.300 kWh/GB × 494 g/kWh = 0.148 g CO₂e per visit

If The Green Web Foundation lists your host as green, the data-centre operational term (0.055 of the 0.300) is removed. That is about 18% of the total, so the same 1 MB page on a listed host comes out at about 0.121 g (0.001 × 0.245 × 494). Our methodology page walks through every step, and our guide to green hosting and carbon ratings explains what a green listing does and doesn't mean.

What the grade means

The letter grade is the SWDM v4 digital carbon rating for one page view: A+ is 0.040 g or less, A up to 0.079 g, B up to 0.145 g, C up to 0.209 g, D up to 0.278 g, E up to 0.359 g, and F above that. The Sustainable Web Design group set those bands from page sizes in the HTTP Archive's June 2023 crawl, so the grade is a quick way to compare one page with another. Notice that our 1 MB example (0.148 g) lands just inside C, while the same page on a listed green host (0.121 g) is a B. You can see every band on the ratings page.

What a carbon test cannot tell you

The model's authors are open about its limits. They describe data transfer as a proxy, note that the model's broad boundaries mean it is likely to overestimate actual emissions, and say it is less suited to pinpointing exactly which part of a system to optimise. In practice, a single test can't tell you:

• Your whole website's footprint. The test covers one page. Your blog, shop and checkout pages may be much lighter or heavier than your home page.
• Your real visitors' experience. Repeat visitors, cached files, different devices and logged-in pages all change what is transferred. Google describes the PageSpeed Insights lab test as a simulated load on a single device with fixed network conditions.
• Your local electricity mix. The global average grid figure is used for every page, so a site whose visitors are mostly on a cleaner or dirtier grid will not see that reflected.
• Server-side work. Heavy database queries, background jobs, email sending or AI features don't show up as page bytes, so they are not in the estimate.
• Your company's carbon footprint. Offices, travel and supply chain are outside the scope. A page estimate is not a corporate carbon report, and it is not a certification.
• Whether your host is carbon neutral. A green-hosting listing means the provider has shared evidence of renewable energy with The Green Web Foundation. It is not proof of carbon neutrality.

One more caution: multiplying the per-visit figure by your monthly visits gives a rough scale, not an inventory. For example, 10,000 visits to a 1 MB page works out at 10,000 × 0.148 g, about 1.5 kg CO₂e. Because every visit is treated as a first, uncached visit, that figure leans high.

How to use your result well

1. Use it to compare, not to declare. The estimate is strongest for comparing two versions of the same page, or your page with a competitor's, because the same rules apply to both.
2. Test the pages that matter. Check your most visited pages and your key journeys, not only the home page.
3. Look at what makes up the weight. Images, video, fonts and third-party scripts are the usual suspects. Our guide to reducing page weight without sacrificing design covers practical fixes.
4. Retest after changes. Results vary a little from run to run, so look for clear changes in page weight rather than tiny shifts.
5. Describe it accurately. If you share a result, say it is an estimate per page view using SWDM v4. Our FAQ explains what the numbers and hosting labels do and don't claim.

Check your page

Run your own page through the free GreenTracer checker to see its measured page weight, estimated emissions per visit, grade and hosting status. Then read how GreenTracer works if you want the full method.

Sources

• Sustainable Web Design: Estimating Digital Emissions (SWDM v4)
• Sustainable Web Design: Digital Carbon Ratings
• The Green Web Foundation, CO2.js: Methodologies for calculating website carbon
• Chrome for Developers: Lighthouse, Avoid enormous network payloads (total byte weight)
• Google: About PageSpeed Insights