What is the Climate Impact of Click Fraud?

Last month, the Telegraph published an investigation into the anatomy of click farms, documenting some of their inner workings – one alarming finding: "as much as half of all internet traffic is fake.” The article also pointed out fake traffic’s impact on both consumers’ perception of popular opinion and the cost to marketers – nearly $100 billion is wasted annually on displaying ads to an audience that doesn't exist!

Beyond the societal and financial damages of bot traffic, how does this behavior impact online advertising’s carbon footprint?

Given what we already know about waste online (there is a strong correlation between things like MFA and high carbon inventory), we conducted our own study to explore the environmental impact of fraudulent fraudulent clicks.

To determine the extent of the climate threat associated with diverse forms of fraud, we partnered with FouAnalytics, a firm that specializes in measuring fraud in ad interactions and overall website activity.

Notably, FouAnalytics is able to identify certain fraudulent practices that remain concealed during routine website inspections, only manifesting when specific codes are embedded in the URL. These mechanisms can result in a surge of additional web requests, inevitably causing increased levels of emissions.

For this study, we utilized a representative sample of 3,024 sites in the US spanning various categories, measuring the presence of fraudulent activity in relation to monthly impression volume and emissions.

Figure 7 above plots fraudulent activity versus emissions by site, where the size of each dot represents the volume of impressions. The bottom-left region indicates low-fraud, low-emissions inventory – approximately 61% of the inventory (fraud <30%, emissions <200 gCO₂PM). The remaining 39% is waste in the ecosystem. Advertising on sites known for fraud is not worthwhile, especially given the outsized emissions.

Scope3 measures the relationship between emissions and fraud by: grams of gCO₂ per 1,000 non-fraudulent impressions, also referred to as effective gCO₂PM (e-gCO₂PM). For example, if a website emits 100 gCO₂PM and has a fake interaction rate of 25%, the resulting effective emissions would be 150 e-gCO₂PM.

Using that metric, the distribution of emissions pulls higher when fraud is filtered out, with a ~33% higher median e-gCO₂PM: from 127 to 168.7. Additionally, the impression-weighted average increases to 245 e-gCO₂PM, indicating that 21.7% of emissions are associated with fake web activity.

The biggest takeaway: Fraud contributes an estimated 353K mt in US programmatic display each year.

Applying this figure to the total emissions associated with US programmatic display (267.8 gCO₂PM, from page 3), the result is that 58.8 grams of CO₂ emissions are wasted on fraud every 1,000 impressions. With an estimated 500B monthly impressions, the estimate is that 352.8k mt of emitted CO₂ is wasted on fraud every year in the US alone.

For more insight into emissions across the digital advertising ecosystem, download the latest State of Sustainable Advertising report here.