You Didn’t Know Bots Solve Captchas, And Do Pharming With HumansSeptember 20, 2020
It is still surprisingly common for marketers to assume that captchas can stop bots from committing various forms of digital fraud, like ad fraud, lead fraud, affiliate fraud, etc. Captchas are those annoying grids that pop up asking you to pick out the fire hydrants, traffic lights, bicycles, and buses. Those are annoying to actual humans, and virtually no deterrent for bots.
This is not even new. In 2013, Business Insider was already reporting on tech that could be used to solve captchas at scale. Today, we see bots solving captchas easily and even hiring low-wage humans to help too. There’s even a helpful YouTube video that shows how it’s done.
Pharming With Humans
The screen shot below shows a Google image search for “chinese bot farms” where racks of mobile devices are connected to power and cooling systems, so that millions of clicks, likes, and shares can be generated continuously in service of whoever pays for clicks, likes, and shares from real mobile devices. The picture further below shows “pharming” (phone + farming) operations where low-wage humans do repetitive tasks like create Facebook, Instagram, Snapchat, and TikTok accounts. These tasks are doled out by algorithms and accepted by humans, much like Uber drivers and Seamless delivery people do when they accept jobs, matched to them by algorithms. There are services you can pay to automate all this for you, so you don’t need any chinese bot farm contacts yourself.
Further reading: Tools of the (Ad Fraud) Trade (Forbes)
Web Traffic Bots – Lots and Lots of Them
Now that you understand bots solve captchas easily and regularly, you may also understand they can easily avoid detection by bot detection tech companies like IAS (Integral Ad Science), DV (Double Verify), and Moat (owned by Oracle). In fact this is evident in the abundance of web traffic sellers that offer “IAS compliant traffic” or “DV compliant traffic.” That means the traffic gets by IAS and DV detection and gets marked as “valid” even though it is not from humans visiting websites. Method Media Intelligence has compiled a gallery of hundreds of Web Traffic Dealers selling all forms of web traffic — 18 gallery pages worth of screen shots.
Ad Fraud – Hugely Profitable Use of Bots
Putting this all together – these bots are used to generate traffic to websites, in order to generate enormous quantities of ad impressions. The bots get by fraud detection, so the ads are accepted and paid for as “valid.” Fraudsters proudly show proof of profits to each other in friendly competition to see how much money they can make off of unsuspecting advertisers, buying through digital programmatic channels, and not directly from mainstream, legitimate publishers. The screen shot below is from a traffic seller dashboard, which shows a 100X return — $211k spent to get $25M payout.
Even “accidental fraudsters” are making money hand over fist. In the following thread on reddit/adops (linked screen shot), this fraudster claims he didn’t know that 99% of his 99% sourced traffic could not possibly be human, but yet were not marked as IVT (invalid traffic) by IAS or DV. So he kept selling the ad impressions and arbitraging a 300 – 800% profit margin. Not as good as the 100X return above, but better than most marketers see from their digital campaigns.
Marketers, look at your own analytics; if something looks strange, investigate it more deeply. You will be able to find the fraud if you looked — you can then turn off the domains and apps that are eating up your budgets, so more of your dollars go to better sites and apps.
Fraud So Obvious, You Can See It in Analytics
Solve Ad Fraud For Yourself With a Pinch of Common Sense