AI creates content, AI edits it, and AI watches it… The “strange cycle” taking shape in “short videos”

Boosting View Counts Through “AI-Driven Viewing”

By now, almost everyone has probably watched a YouTube Short. They provide instant gratification in a very short amount of time, and since the next video starts playing immediately, you end up spending a considerable amount of time on them before you realize it. Just like Instagram Reels and TikTok, Shorts have become one of the things that consume a massive amount of time for people today.

So-called “short videos” are often created by extracting only the parts of a longer original video that are likely to attract the most viewers.

While these short clips are often created to drive traffic to the original YouTube video or similar content, sometimes their sole purpose is to inflate metrics like view counts and subscriber numbers. Recently, it has become easy to create videos using AI and even have AI dub the audio. Furthermore, there are now examples of people “getting AI to watch” their content just to boost view counts.

In 2024, a man was arrested on charges including telecommunications fraud after having AI bot accounts play hundreds of millions of AI-generated music tracks he had uploaded billions of times, earning approximately 1.4 billion yen in royalties. There are reportedly other cases where video view counts are being inflated using similar tactics.

“AI-driven view inflation is a mechanism that increases video view counts by issuing instructions to ‘View Bots’ and other tools that automate viewing behavior.Bots model human viewing behavior, and in recent years, they have also begun to mimic actions such as ‘watch time,’ ‘where the video is paused,’ ‘seek bar manipulation,’ and ‘scroll speed and click intervals.’

Since platforms like YouTube are taking measures to combat fraudulent activity, these bots are able to circumvent those measures and inflate view counts by mimicking human behavior more closely.

“In recent years, some methods have also incorporated browser automation—a technology that uses programs to replicate and automatically execute the actions a human would perform in a browser. It has evolved to the point where it can perform a series of actions such as ‘opening a browser and searching for videos’ and ‘watching related videos,’” says an IT writer.

It appears that the platforms’ countermeasures against such fraudulent view count inflation are also evolving daily.

“They analyze user sessions and determine whether they are valid based on various factors, including device, network, and behavioral analysis. Since simply acting ‘human-like’ is enough to get detected, these operators are developing even more sophisticated techniques to evade monitoring. It’s a never-ending game of cat and mouse,” (ibid.)

Methods Spread Through “Production Manuals”

Short videos produced using AI typically draw on content that has gone viral on social media, and they are often created using videos from overseas. Recently, around the end of December last year, a video titled “Former Child Actor Who Starred in a Popular Overseas Drama Now Living on the Streets” went viral on social media and sparked widespread discussion.

This led to the creation of explanatory videos on platforms like YouTube and Instagram, resulting in widespread “reuse of viral content.”

Common topics for these short videos include everyday questions and curious phenomena—things that grab people’s attention. While AI is used to create explanatory videos based on these clips, the explanations themselves are often incorrect. Content that can hardly be considered legitimate is being churned out and flooded onto YouTube.

Why have these types of short videos spread so rapidly? When we spoke with Yusuke Yamano, a writer knowledgeable about AI editing of images and videos , he pointed out that “one factor is that the production methods spread rapidly as information products on platforms like note.”

“For example, if you search for ‘AI short videos’ on note, you’ll find plenty of step-by-step guides—ranging from free to paid content. They explain things like, ‘Use ChatGPT to write the script and AI-generated voice for the narration.’“The creators’ objectives vary widely—from attracting an audience through short videos to recruiting people for LINE groups or their own paid note accounts,” says Yamano.

Why are videos from overseas the ones circulating as short videos? One reason, according to Mr. Yamano, is that “it’s difficult for rights holders—such as the video’s creators or uploaders—to discover them.”

“Since these short video clips are generally created without permission, there’s a risk of receiving a takedown notice if they’re discovered. If they’re edited for a Japanese audience, YouTube’s algorithm makes it harder for them to reach people overseas. The fact that it’s relatively safe to create pirated versions has spread, and this has likely become the standard.“There are a massive number of clips from Japanese TV shows circulating overseas,” (Mr. Yamano, same below)

How to Make Money by Selling Accounts

Apparently, the surest way to make money from mass-produced short videos is to sell them on websites dedicated to buying and selling YouTube accounts.

“There are numerous sites where you can buy and sell YouTube accounts, such as ‘M&A Club’ and ‘Rakko M&A.’ The selling price is determined by factors like the number of subscribers and the type of content the account features. Even accounts with fewer than 10,000 subscribers can sometimes be traded for between 100,000 and 200,000 yen.

The key factor is how much revenue can be expected after purchasing the account. For videos featuring human speakers, a change in the performer following a sale could lead to a drop in views and a decrease in revenue. For this reason, accounts that use AI voices to minimize reliance on specific individuals are preferred.”

Everything from scriptwriting to editing can be done entirely by AI. There are even imitators who create videos that are exact copies of viral short videos. Mr. Yamano adds , “There are people selling scripts for AI creators, but those are also generated by AI. They feed AI-generated content into AI to create videos. It’s a vicious cycle, so the result is nothing but low-quality videos.”

Every day, videos are uploaded online solely to pad view counts. Videos are created using AI to generate numbers; those videos are then edited into short clips by AI; and in the worst cases, bots even “watch” the videos to inflate the view counts.From production and editing to automated viewing, a strange cycle is emerging in which “AI-generated content is consumed by AI” without any human intervention.

At the “13th Nikkei ‘Hoshi Shinichi Award’” announced in February, three of the four winning works in the general category—including the Grand Prix winner—were created using generative AI, leading judges to remark that “it was impossible to tell the difference.”

The distinction between what AI creates and what humans create is becoming increasingly blurred. While finding the right way to coexist with AI remains a major challenge for the future, the “contamination” caused by AI-generated content is already spreading widely across the internet.