M MurmTools

Amazon CloudFront in the Consent Gap Index

Amazon CloudFront is Amazon's content delivery network. Sites use it to serve the heavy, static parts of a page — images, video, fonts, JavaScript bundles — from a server near the visitor instead of from one origin machine. You meet the name in a request log, or in a content blocker's list of third parties, where it looks alarming and mostly is not. This index recorded it on 349 sites, 6.6 percent of the 5,294 measured, which makes it the 6th most-encountered of the 88 companies here. It is not classified as tracking.

A delivery network, not a data collector

The index carries a category for CloudFront — CDN — and a classification: not tracking. That distinction matters more here than on most entries, because CloudFront's high site count is easy to misread. Appearing on 349 sites does not mean 349 sites hired Amazon to observe their visitors. It means 349 sites host files on AWS, or use a product that does. A CDN request carries the visitor's IP address and user agent to the delivery network because that is how a request works, not because anyone chose to send it. The interesting question about CloudFront is therefore not what it does, but what tends to accompany it.

What else is on the page

Google was present on 269 of the 349 CloudFront sites, 77.1 percent — the single most common companion in this entity's own co-presence figures. OneTrust follows on 86 sites (24.6 percent), jsDelivr on 76 (21.8), Cloudflare on 47 (13.5), Cloudflare Analytics on 40 (11.5), Meta on 37 (10.6), HubSpot on 33 (9.5) and Google Tag Manager on 32 (9.2). These are counts of sites where both appear, not a claim that one loaded the other. Read as a picture of a market rather than a chain of causation, it is a fair sketch of how a well-resourced site is assembled in practice: AWS underneath, Google measuring, a CDN or two for libraries, a consent vendor on top.

The median CloudFront page

9 distinct companies, 2 of them classified as trackers, across 59 requests to third-party hosts. Among the 88 companies in this index, that request count ranks 20th and the company count 43rd — a page that talks a lot to a mid-sized cast. First contact on the median site landed at 653 milliseconds, ranking 51st of 88; the fastest recorded on any CloudFront site was 91 milliseconds. Early contact happened on 326 of the 349 sites, 93.4 percent. The median site had 7 companies contacted before an answer could be registered, and none after the consent window closed.

Banners on just over half

182 of the 349 sites carried a consent tool. Six were seen alongside them: OneTrust on 96 sites, an IAB TCF consent tool on 45, Cookiebot on 13, Osano on 11, CookieYes on 7 and Usercentrics on 6. Those figures are counts of sightings rather than slices of the 182 — a site can carry more than one banner, and some carry none of the six by name. What the split shows plainly enough is that CloudFront's population is bimodal: roughly half of it is European-facing enough to need a dialog, and roughly half is not.

Twelve sites by name

discord.gg contacted 7 companies, all 7 early, at 325 ms; discord.com the same 7 of 7 at 524 ms. flickr.com shows 8 of 8 early at 457 ms behind an Osano banner, indeed.com 4 of 4 at 693 ms and duolingo.com 3 of 3 at 1,632 ms, both behind OneTrust. Heavier and more deferred: slack.com contacted 10 companies with 8 early at 806 ms; salesforce.com 9 companies but only 4 early, with first contact at 1,742 ms; ibm.com 9 with 5 early at 528 ms. eventbrite.com shows 8 with 6 early at 411 ms, media.net 8 of 8 at 745 ms, smartadserver.com 7 of 7 at 667 ms and opendns.com 6 with 5 early at 1,013 ms.

What to do with a CDN entry

If you are auditing a page and CloudFront shows up, the useful next step is not to remove it but to look at the other eight companies the median CloudFront page contacts. The delivery network is the least interesting name on that list and usually the easiest one to explain. The measurement method and the definition of the early window are at /consent-gap; if you want the same breakdown for a page in front of you rather than one of the 5,294 here, the MurmStack extension produces it live.

Highest-ranked sites where this was seen
SiteCompaniesBefore consentFirst contact
discord.gg77325 ms
discord.com77524 ms
flickr.com88457 ms
ibm.com95528 ms
salesforce.com941742 ms
media.net88745 ms
smartadserver.com77667 ms
opendns.com651013 ms
slack.com108806 ms
indeed.com44693 ms
eventbrite.com86411 ms
duolingo.com331632 ms
Consent tools seen on these sites
ToolSites
OneTrust96
IAB TCF consent tool45
Cookiebot13
Osano11
CookieYes7
Usercentrics6

The six most common are listed. The remaining 4 of 182 sites with a banner carry something outside that six. One tool is recorded per site, so these counts never overlap.

Seen on the same pages

Companies most often contacted by the same sites. Co-presence only — the crawl records what arrived, not who called whom.

Questions

Is Amazon CloudFront a tracker?

It is classified in this index as a CDN and not as tracking. It receives the network-level details any server receives when it is asked for a file. That is different from a script installed to record behaviour, which is what the tracker classification marks.

Why does it appear on so many sites?

349 of the 5,294 measured sites, 6th of the 88 companies indexed, because a large share of the web hosts assets on AWS or uses software that does. Site count for infrastructure reflects hosting decisions rather than measurement decisions.

What does the 77.1 percent Google figure mean?

Google was present on 269 of the 349 sites where CloudFront was present. It is a count of sites where both appear. It does not say one requested the other or that either arrived first.

How many CloudFront sites showed a consent banner?

182 of 349. OneTrust was the tool most often seen alongside them, on 96 sites, followed by an IAB TCF tool on 45.

Check any page yourself MurmStack is the tool that produced these figures. Free, two permissions, no host access.
Add to Chrome — free