Fake YouTube Views Still Monetized by Google

Google charges advertisers when their YouTube ads are viewed by humans or robots, according to new research from several European research institutions.

The study was conducted by seven researchers from NEC Labs Europe; IMDEA Networks Institute; Politcenino di Torino, an Italian engineering university; and Universidad Carlos III de Madrid, a university in Spain. It examined the fraudulent view detection tools of five major portals: YouTube, Dailymotion, Vimeo, MyVideo’s German domain, and TV UOL, one of the most visited websites in Brazil. According to Understanding the Detection of Fake View Fraud in Video Content Portals, YouTube’s detection systems are by far the most sophisticated in every area, identifying 83 percent of the bots the researchers unleashed on their own videos. However, 91 percent of the views were charged by AdWords.


In other words, Google was able to recognize these views as fake, discounting them from the public view count, but still monetized them. To determine this, the research team employed several strategies, careful not to be too careful.

“We leverage the modularity of the probe to define different configurations to isolate the impact of different parameters on the fake view detection mechanism,” the study says. “In our experiments, each probe instance uses a single public IP address chosen from our pool and performs 20 views per day to the same video for eight consecutive days. We expect that this behavior is aggressive enough to trigger the fake view detection mechanism of YouTube.”

Parameters used include setting consistent view times to eliminate randomness, varying the view bursts and inter-view wait times, using the same cookie to gauge what role they play in the detection, and consecutive short views that are atypical for real users. In addition, they set out to emulate some real people behaviors, such as setting the browser and referral site for sessions.

The latter was YouTube’s biggest hurdle in detecting fake views, though the video platform was onto 98 percent of view bursts: 20 consecutive views over a period of 24 hours from the same IP address. The other methods were also mostly shut-down, leading researchers to determine that variability is a big factor in successful bot attacks. Also, they found that YouTube is more susceptible to attacks that employ many IP addresses.

Examining four different videos, the study found that the two using single IP addresses had public view counts of 18 and seven, and monetized view counts of 31 and 60, respectively. The one with significantly more fake views charged had twice as many daily views. A third video, which had 70 daily views on two IPs, charged for 15 views despite not counting a single one. On the other hand, the video with 10 views on a single IP monetized 178 views, but still counted 147 of them.


“YouTube uses a seemingly permissive detection mechanism to discount fake monetized views. This exposes advertisers to the risk of building their advertisement campaigns on unreliable statistics, and may make them initially burden the risk of fraud,” the summary says, adding that the detection of fake monetized views results in a severe penalty from Google.

These numbers are particularly alarming, when compared with those from the Social Media Examiner’s Social Media Marketing Industry Report from earlier this year. That study found that 57 percent of marketers currently use video and 72 percent interested in it plan on increasing their use, particularly on YouTube.

A Google spokesperson told Financial Times that the search giant will contact the researchers to discuss their findings, assuring that the company has invested heavily in technology and talent to “keep this out of our systems. The vast majority of invalid traffic is filtered from our systems before advertisers are ever charged.”

Related reading

trint makes spoken words videos crawlable
Eight tips to getting your videos ranked on YouTube in 2019
youtube and child safety: is the service doing enough?