A Scoping Review of Methods to Measure the Energy and Carbon Footprint of Web Tracking and Advertising

Every time you visit a website, a hidden economy of data exchange happens in the background. While you read an article or watch a video, invisible scripts are running to track your behavior, build a profile of your interests, and deliver targeted advertisements. This system, often called the ad-tech ecosystem, is the primary way many free online services make money. However, capturing, storing, and processing this massive amount of user data requires significant computing power. That power comes from electricity, and the generation of that electricity often produces carbon emissions. As the internet grows, so does its environmental footprint, raising a difficult question for researchers: how much of that energy is actually spent just on the ads and the tracking, rather than on the content you are trying to see?

A team of researchers at the University of Toronto set out to answer this by reviewing the existing scientific literature. They were not conducting a new experiment themselves but were instead mapping the landscape of how other scientists have tried to measure this specific environmental cost. They found that the field is still young and fragmented. The studies they reviewed are scattered across different disciplines, using different words to describe the same processes and different tools to measure the same things. To make sense of this, the authors gathered 15 key studies from an initial pool of 46 candidates and organized them into five distinct methods that researchers use to isolate the energy cost of web tracking and advertising.

The most common approach, used in the majority of the studies, is called ad blocking. In this method, researchers visit a list of popular websites twice: once with a standard browser and once with an ad blocker turned on. By comparing the energy used in both scenarios, they can calculate the difference caused specifically by the ads and trackers. Some researchers use browser extensions to block the ads, while others use built-in browser features or even block the requests at the network level. To get these numbers, they often use software that reads the computer's processor usage or battery drain, or in some cases, they connect a physical voltmeter to the device to measure the electricity flowing into the hardware. This method is practical because it can be done on a regular computer, but it relies on the assumption that the only thing changing between the two visits is the presence of the ads.

A second group of researchers takes a more controlled approach by recreating websites in a laboratory setting. Instead of visiting live sites on the open internet, they download the pages and strip away the ads and tracking scripts themselves, creating a clean version of the site. They then serve these versions locally on a server they control, ensuring that no outside variables interfere with the test. In one variation of this method, researchers built a website from scratch and added different numbers of tracking tools to it, watching how the energy consumption rose with each addition. This offers a high degree of control over the experiment, allowing scientists to see exactly how much energy a single tracker consumes, though it may not perfectly reflect the chaotic reality of the live web.

A third, more unusual method involves replaying individual advertisements. One study collected thousands of ads from the web and created a system to display them one by one on a blank browser page. By measuring the energy required to load a single ad against a baseline of an empty page, the researchers could determine the cost of rendering that specific advertisement. They then used this data to train a computer model that could predict the energy cost of other ads based on their characteristics. This technique allows for a granular look at the cost of individual ad units, but it requires a massive amount of manual data collection to build the initial library of ads.

The fourth approach looks at the flow of data itself rather than the device displaying it. Instead of measuring the computer screen, these researchers analyze the network traffic moving between the user and the internet. They examine the data packets sent and received, trying to identify which parts of the data stream belong to advertisements and tracking cookies. By summing up the total volume of this specific traffic, they can estimate the energy required to move that data across the internet. However, the authors of this review note a significant flaw in this method: simply knowing how much data is transferred does not automatically tell you how much energy was used to process it, as the relationship between data size and energy consumption is complex and not yet fully understood.

The final method relies entirely on existing numbers rather than new measurements. Researchers using this approach take figures and coefficients from other published studies—such as the average energy cost of moving a gigabyte of data or the carbon intensity of the power grid—and apply them to estimates of how much data advertising generates. This allows for broad, large-scale calculations without needing to run any experiments or collect new data. While this is the most accessible way to get a rough estimate, its accuracy depends entirely on the quality and relevance of the original numbers it borrows.

The review reveals that the field is just beginning to find its footing. Eighty percent of the studies they analyzed were published in the last five years, indicating a rapid surge in interest as the environmental crisis deepens. Despite this growth, the researchers found that the studies rarely talk to one another. Different teams use different terms for the same concepts, and they often fail to build upon each other's findings. For instance, one study found that the size of the data transferred for an ad does not necessarily predict the energy needed to display it, a finding that challenges the assumptions made by other studies that rely on data volume to estimate energy use.

Currently, there is no single standard method that everyone agrees on. Most of the research focuses on what happens on the user's device, leaving a large gap in our understanding of the energy consumed by the massive data centers that store and process this information. Because these corporate infrastructures are opaque and difficult to access, researchers have had to rely on indirect methods or controlled simulations. The authors suggest that future work needs to develop better ways to measure the energy used in these data centers and to create a shared language so that scientists from different fields can compare their results accurately. Until then, we have a patchwork of methods that offer glimpses into the environmental cost of the digital economy, but no complete picture.