Most business websites collect pages over the years: an old event, a product you no longer sell, a news post from 2019. Once you start thinking about your site's carbon footprint, it's natural to wonder whether those forgotten pages are quietly adding to it. The honest answer has two parts. Under the model GreenTracer uses, a page nobody visits adds nothing to the per-visit estimate. But that estimate doesn't describe everything your website does, and knowing the difference helps you put your effort where it counts.
What a per-view estimate counts
GreenTracer uses the Sustainable Web Design Model v4 (SWDM v4). Its main input is data transfer, which the model defines as the amount of data transferred when loading or using a digital product or service, and its headline output is average emissions per page view. It works out an energy figure for each gigabyte transferred across three parts of the system: data centres, networks and the devices people use, counting both the electricity they use and the emissions from making the equipment.
In practice that means the estimate scales with two things: how much data a page sends, and how often it is sent. Using the model's figures, each megabyte transferred works out at about 0.148 g CO₂e per page view on a standard host. As an illustration, a 2 MB page comes to roughly 0.30 g per view. Viewed 1,000 times, the same arithmetic gives roughly 300 g. Viewed no times at all, it gives nothing.
So what about a page with no visitors?
If nobody loads a page, no data is transferred for that page, so under SWDM v4 it contributes nothing to the per-view figure. That's also why a GreenTracer check describes one page view of one page, rather than adding up every page you've ever published.
Be careful with “nobody”, though. A page with no views in your analytics may still be loaded. If your analytics only run after a visitor accepts cookies, people who decline won't be counted (see cookie banners, tags and your carbon test). Treat a low count as “rarely visited” rather than “never visited”.
What the per-view figure doesn't cover
An unvisited page still exists. Its files sit on a server, and probably in your backups too. It's worth being precise about how SWDM v4 treats that, because it's easy to overstate in either direction.
- The model is top-down. Its authors describe SWDM as an attributional model: it takes estimates of the energy used by each part of the system and divides them by the data transferred. The data centre part is described as the energy required to house and serve data, so that energy is spread across transferred data rather than given to individual stored pages.
- It doesn't estimate your storage, backups or build tools separately. The model's page doesn't give a way to estimate the footprint of keeping a particular page stored, of backing up your site, or of the systems that build and deploy it. The authors also say it will not help you estimate emissions for individual parts of those systems.
- It's an estimate. The authors say models and estimates are imprecise by nature, and that the model is likely to overestimate actual emissions.
So the fair way to put it is: an unvisited page adds nothing to your per-visit estimate, and the per-visit model isn't designed to tell you what storing it costs. We won't put a number on that, because the model doesn't give one.
Find the pages people actually visit
Because the per-view estimate is multiplied by visits, the pages that matter most are the ones people load most. Your analytics tool will tell you which those are. In Google Analytics 4, the Pages and screens report (Reports, then Engagement, then Pages and screens) lists your pages with a Views count, which Google defines as the number of web pages your users saw. Other analytics tools have an equivalent “top pages” view.
- Choose a period long enough to smooth out quiet weeks, such as the last three months.
- Sort by views and note your top five to ten pages.
- Check each of those pages and note its page weight and grade. The rating page explains what each grade band means.
- Put a heavy page with lots of views at the top of your list.
Where to spend your effort
Saving a megabyte on a page seen a thousand times a month is worth far more, under the model, than saving the same megabyte on a page seen twice. Start with the pages at the top of your analytics list, and with the shared parts every page loads, such as your header images, fonts and scripts, because a saving there applies to every view. The quickest wins are usually images and web fonts.
What about the old pages? Tidying them up is still worthwhile, but mostly for your visitors: out-of-date prices or events can confuse people, and a smaller site is easier to keep accurate. If you remove a page that others link to, point its address at the most relevant current page so those links still work. Just don't expect deleting unvisited pages to change your per-view estimate. Spend that time on the pages people actually see, then check them again after each change.
