Most business websites load code from other companies: a tag manager, analytics, a chat widget, a booking tool, video players, fonts. A website carbon test counts these files like any others, because they are data your visitors download. But there is a twist. If your site waits for cookie consent before loading some of them, a carbon test that never clicks the banner won't see them. This article explains what that means for your result and how to check what your tags really add.
What happens before anyone clicks
GreenTracer measures the page with Google PageSpeed Insights, which loads it like a first-time visitor and does not interact with it. It doesn't scroll, click or accept a cookie banner. So the result depends on how your site handles consent:
- Tags that load regardless of consent are counted in full. Some consent set-ups load the tag manager and analytics straight away and only limit what they store. Those files are still downloaded.
- Tags that wait for consent are not counted, because the test never gives it.
- The banner itself, its script and any styles or fonts it brings, is counted.
We checked this on GreenTracer itself. Before October 2026 our pages loaded Google Tag Manager before a choice was made. It respected the “not allowed” setting for cookies, but it still made requests. We changed the site so the tag manager only loads after a visitor accepts analytics. Before a choice, or after “Reject”, the page now makes no request to Google's tag or analytics servers at all.
Two loads worth knowing about
Because of this, your page effectively has two weights: one for visitors who haven't accepted (which is what the carbon test sees) and one for visitors who accept. If most of your visitors accept, the second matters more than your test result suggests. Neither is wrong; they describe different visits. It's worth knowing both before you decide what to change.
Audit your tags in ten minutes
- Open your page in a private window in Chrome, so you start without a saved choice. Open DevTools (F12) and the Network panel, tick “Disable cache” and reload.
- Note the total transferred at the bottom of the panel. Then use the filter box with a minus sign to hide your own domain (for example
-domain:example.com). What remains is third-party. - Accept the cookie banner, reload, and note both figures again.
- Compare. Each third-party host you see should map to something you know you use. Lighthouse's third-party summary lists the same hosts with their transfer sizes.
Trim without losing what you need
- Remove tags nobody uses. Old pixels and trials are common. If you can't say who reads a tag's data, it's a candidate for removal.
- Load embeds on demand. A video player or map can show a lightweight preview and load the real embed only when someone clicks it. Google calls this a facade.
- Gate optional tags on consent. If a tag needs consent, don't load it until it has consent. That is lighter for people who decline, and it's cleaner for privacy. Check the set-up with your consent tool and Google's consent guidance if you use Google tags.
- Keep fonts lean. Fewer families and weights, and only the character sets you need.
Every megabyte removed from the load lowers the estimate by about 0.148 g CO₂e per page view under the Sustainable Web Design Model v4 (about 0.121 g on a host listed as green), and the saving applies to every visit that would have downloaded it.
Lighter and more private go together
Fewer third-party requests mean less data for visitors to download and fewer companies receiving information about each visit. Neither goal has to come at the cost of the tools you actually rely on. Start with the audit, remove what you don't use, and load the rest only when it's needed.
Check your page, then open “How we estimated this” in the report to see what the test included.
