Google Employs 'Hundreds' Of Humans To Sniff Out Ad Fraud On The Web

Botnets generate clicks automatically, at the same overall rates as humans, making them difficult to detect. Some publishers claim, and bill for, vast audiences even though few people in real life seem to have heard of their publications. And some web sites serve ads "below the fold," meaning beneath the area of the site displayed on your screen — charging clients for impressions that people can't see.

One company, Spider.io, estimates that advertisers waste $6 million a month on fraudulent transactions, and that 80% of impressions delivered to some brand campaigns can be false.

Advertisers can't see the fraud because they use automated, "programmatic" ad buying software on real-time bidding platforms. They're just buying numbers, in other words.

Yet Google, AdExchanger notes, is always "among the top inventory platforms, [and] Google is widely regarded as the standard-bearer of policing ad fraud."

The company has a "secret" method for weeding out fraud, it says in a blog post. "Human review":

We have hundreds of employees around the world who review web pages, test our partners’ downloadable software, and prevent ads from showing on sites that violate our policies. Depending on the severity and persistence of the offense, they may stop ad serving on that page or site, or across the publisher’s entire account.

Apparently, there is still a role for humans in automated ad buying — fixing the robots' mistakes.