A GreenTracer report opens with three boxes: emissions per page view, a carbon grade and a hosting status. They sit side by side, but they answer three different questions. The estimate says roughly how much carbon one visit to the page is associated with. The grade puts that estimate in a band so you can compare pages. The hosting status says whether your host appears on a public green-energy list. Reading them separately is the easiest way to avoid over-reading any one of them.
The estimate: grams per page view
The headline figure, for example “0.236 g CO₂e”, is an estimate for one view of one page. It is not a meter reading. Nobody can observe the electricity used by the servers, networks and devices involved in a single visit. What can be measured is how much data the page transfers, and the Sustainable Web Design Model (SWDM) v4 turns data into energy and energy into carbon.
GreenTracer applies the model's published constants: 0.300 kWh per gigabyte across data centres, networks and devices (including the energy used to manufacture them), and a global average grid intensity of 494 g CO₂e per kWh. A page that transfers 1 MB (0.001 GB) is therefore estimated at 0.001 × 0.300 × 494 = about 0.148 g. The relationship is proportional, so halving the data halves the estimate.
The report shows the working in “How we estimated this”: the bytes measured, each part of the calculation and the model version. If a figure ever looks surprising, that section lets you check it yourself.
What the test includes and leaves out
The data comes from a Google PageSpeed Insights (Lighthouse) test of the page's initial load by a first-time visitor with nothing cached. That has a few consequences worth knowing:
- Everything the page downloads while loading is counted, compressed as sent, including files from other companies' servers such as tag managers, fonts and embeds.
- The test doesn't scroll or click. Images set to load lazily further down the page are usually not requested, and anything that only appears after a click isn't counted.
- The test doesn't accept cookie banners, so tags that wait for consent don't load.
- Repeat visitors usually transfer less, because their browser has cached files. GreenTracer doesn't discount for that, so the estimate leans towards a first visit.
Mobile, desktop and the heavier result
Each check loads the page twice: once as an emulated mobile phone (about 412 pixels wide) and once as a desktop browser (about 1,350 pixels wide). The two can differ a lot, mainly because responsive images send larger files to wider screens. Your report shows both, labelled by device, and uses the heavier of the two for the headline figure and grade.
Using the heavier result is GreenTracer's own cautious policy. The SWDM doesn't require it, and it can mean a lower grade than a phone-only test would give. If your mobile and desktop grades differ, the page is probably sending large images to desktop screens; our article on why desktop and mobile results differ covers how to check.
The grade: a band, not a verdict
Grades use the SWDM v4 digital carbon ratings. Each grade covers a range of grams per page view, and the upper limit belongs to that grade:
| Grade | g CO₂e per page view |
|---|---|
| A+ | up to 0.040 |
| A | over 0.040, up to 0.079 |
| B | over 0.079, up to 0.145 |
| C | over 0.145, up to 0.209 |
| D | over 0.209, up to 0.278 |
| E | over 0.278, up to 0.359 |
| F | over 0.359 |
A grade is a convenient way to compare pages and track progress. It isn't a certificate, and the letter doesn't say anything the estimate doesn't. A page just over a boundary and a page just under it are nearly identical, which is why the report always shows the grams as well.
Hosting status: a listing, not a guarantee
GreenTracer asks The Green Web Foundation whether your host is listed as running on green energy. A listing means the provider has shared evidence with them. It is not proof that your site is carbon neutral, and “not listed” doesn't prove a host uses no renewable energy.
The listing does change the estimate in one specific way: for a listed host, the model removes the data centres' running energy, so the rate falls from 0.300 to 0.245 kWh per gigabyte. Networks, devices and manufacturing are still counted.
Checking again and reading the history
Each report keeps earlier checks of the same page. A repeat check of an unchanged page is usually within a few per cent, because live pages vary: ads rotate, content changes and third-party files differ from one load to the next. If the history shows a “reused measurement”, PageSpeed Insights returned a run it had already taken, so the two checks aren't independent. A real change in your page shows up as a lasting change in the bytes measured, not just one different figure.
What to do with your report
- Start with the “Three useful actions”. Each one is based on something the test measured on your page.
- Treat each potential saving as an upper bound: unused code may still be needed after a click or scroll.
- Check the same page again after a change, and compare the bytes measured as well as the grade.
- If you share the result, share the report link so others can see the date, the scope and the working.
Check a page to see your own report.
