Molloy DSG Professional Digest

Context manipulation, vote brigading, and AI driven bots are a reality of our current internet landscape. These methods are virtually undetectable to an average user, and often associated with bad actors gaming a system. Let’s take a closer look at why YouTube has bots in the first place, and shed light on operating principles that would require a company to take no action for over four years in terms of creating a solution to remove them. YouTube has been infested with bots for a long time prior to this writing. Certain popular content has significant bot activity, while other content depends on bot-like actions. Why is that? There are several types of bot and bot-like activites, and other insidious issues, that plague the YouTube platform and others. Let’s take a look at some basic theories.

• Autoplay-next user botting: Videos that are left unattended by real users are an opportunity to increase view counts for a channel that is advantaged algorithmically to align with YouTube’s always-on-by-default autoplay feature, which instantly loads the next video in the algorithm queue presented to the user unless the end user has that feature switched off. This mechanism provides a mechanism to inflate viewcounts of videos in the next-up position using real users; this becomes the most clear when mainstream network television news segments and talk shows appear up next in this queue amongst completely unrelated content.

  Certain channels are high value advertising targets, presenting an economically-driven incentive to underhandedly insure content from certain channels is displayed even to nobody watching. Stopping either autoplay or providing genuinely related content would inhibit the opportunity to inflate view count, and reduces ad performance data provided to advertisers who pay Google for advertising to these impressions. There’s no way to see or measure phantom impressions on either YouTube itself or Google; as if by design, this enables this practice to continue without scrutiny from their advertisers or viewers.

• View-count botting is almost easy to confuse as the same to the non-technical user, but actual view-botting is much harder to do without significant resources. View-botting is the practice of using bot farms to increase view counts, much like how bot farms are used to create spam comments based off real user comments. It’s not immediately easy to discern if Google prohibits view-botting, but it’s clear comment-botting is allowed.

• Comment-botting occurs when a bot system scans existing user comments for popular text, and creates a comment extremely close or identical to the original. The new comment is then boosted by a bot-like program which harnesses a significant amount of pirate nodes to raise the vote count on said comment. What makes these believable is when this practice is combined with the automatic duplication of usernames and profile pictures to make the offending comment more human.

• SEO-botting is when the comment bot systems use a unique identifier or slogan present in the language used to spam the site. This is often in the form of “I can’t believe he finally had him on! He will finally lead us to XNERDLOL which will save us all!"

  These comments depend on the user to seek out the unknown slogan or word in a search engine, then following the results ultimately to either a financial scam or malware destination. It’s a way to link to a site in YouTube without providing an actual URL.

• Combining comment-botting activity with SEO-botting is especially insidious, as it creates the illusion that the comments on an affected popular video cannot be trusted. If the narrative is that “Russian Bots” are “influencing” something, then a comment-SEO-bot system can achieve the effect of promoting this narrative by stealing real posts, upvoting them considerably at once to game the Top Replies list, then the comment can be edited to include the XNERDLOL search keyword. The objective of utilizing these botting methods provides a malevalent actor to paint a narrative around the subject of the content. When the Top Replies list all has extremely high upvote acounts and SEO spam, it’s an instant red flag for any unfamiliar user – and should be considered an underhanded and subtle manipulation tactic. "I knew it! It's the Russians!!"

Why would a platform allow these practices to continue? For one: money; advertisers pay significantly into YouTube and Google’s operating budget, and it would be a conflict of interest to their finances if popular content were not as popular (whether real interactions, views boosted by autoplay-next feature, or bots.) Another reason is for power; at the very top of any company this large, there’s significant financial influence and deals that they have no obligation to disclose to the user, and allowing video and comment content to be promoted underhandedly can serve these objectives. Third take, is that Google may not actually have the engineering resources available to actually solve the problem. YouTube and Google have become stagnant in their talent, and these problems are hard to solve. It’s clear the talent is lacking based on the quality of their website. Otherwise, if they do have the resources to solve the bot problem, are they then complicit in deception?

It’s important to understand, that a platform that participates in the countless cases of shadowbanning of real users while allowing fraudulent material to remain is a sign of poor platform health. Unfortunately we have to deal with the underhanded tactics of the Google and YouTube system, and how they screw their advertisers with their black-box ad budget demands while simultaneously allowing bots to game their system. It’s a sign that there is room for competition, and much like Google’s failure at deploying a reasonable AI assistant, it’s also an indication that YouTube can be next. With their content manipulation and black-box algorithms, it’s uncertain why YouTube hasn’t been seen as a publisher rather than a platform. Certainly by providing their servies via an in-house algorithm that determines what content gets to the top, which is computed differently for almost every user, can’t simply keep them in the platform categorization forever.