YouTube network graph of 6,902 AI-generated news videos

đź“şFakeTube: AI-Generated News on YouTube

Jonathan Albright
7 min readJan 17, 2017

One of the results of my recent post-election “news ecosystem” effort was the presence of YouTube. Beyond YouTube’s role in hosting videos through embeds on political websites, after reading a piece on A.I. and YouTube by Guillaume Chaslot, I thought I’d look more into why it increasingly feels YouTube, like Google, is being harnessed as a powerful agenda booster for certain political voices.

During this research effort, I discovered an unusual video titled: “A Tease: netanyahu russia jerusalem.”

A Tease?

“A Tease: netanyahu russia jerusalem”

What made this video interesting was the text shown below in the description, which read:

“This Tease is generated using T, one of the most advanced artificial intelligence technology platforms ever created. It includes relevant Thoughts such as: …”

“A Tease…” video description

The “A Tease: netanyahu russia rerusalem” video’s “up next” playlist consisted of dozens of similar, keyword-stuffed politics and current affairs “A Tease…” news videos:

Accompanying playlist for “A Tease: netanyahu russiajerusalem” video: https://youtube.com/UC73Q266ChQ6P9NY_vF0bk4A

Is this the YouTube “fake news” jackpot? Seeing the playlist, and noting the “generated with AI” description, I was curious about the scale of this “auto-generated” news on Youtube. Through a video title-focused API search query, I found 6,902 more “A Tease…” YouTube videos.

YouTube API: Search

Here are some examples of the repetitive yet very time-sensitive video titles:

Examples of 6,902 related videos from “A Tease…” search on YouTube (API)

Below is a close-up of the network graph of these 6,902 AI-generated news videos across 19 YouTube channels:

Network graph of 6,902 related videos from “A Tease…” search on YouTube (API)

19 Channels, 78,349 Videos

In the first search round, I found 19 YouTube channels containing videos titled “A Tease….” The IDs and names of each of these channels are listed below. The channel names include: TipTopBest, A T User, Tip Top, T Drinker 2465, Tiptop Insights, Another T Drinker, Another TipTop User, T Drinker 7438, T Drinker 46895, T Drinker 865110, TipTop T, A TipTop Drinker, T Drinker 6593, T Drinker 743339, T Drinker 176, T Drinker 247, and three “non-T” channels: [Chinese], CNN BRK, and Trump VS Putin.

19 YouTube channels with AI-generated news “Tease” videos

Through a network analysis and channel detail API request, I found the 19 YouTube channels account for 78,349 similarly named, keyword-packed “A Tease…” videos. As you can see in the list above, this is what amounts to a large-scale “fake news” video content farm. YouTube is a big place, but ~80,000 videos is a large number by any measure. Especially given that it’s only 19 channels.

Curated, Republished, and AI-Narrated

Each “T” AI-generated video consists of a progressive “slideshow” of still images related to the title of the video, which originate from various websites, Wordpress blogs, and content delivery networks across the internet. It’s notable that the last frame of these “Tease” videos seem to have a “copyright” disclaimer.

Shows the final frame in “A Tease…” video

A computerized voice “reads” out text from published news articles and other news-related content as the slideshow progresses. The narration is surprisingly accurate and coherent, except for the brief pauses during the transitions between the sources from which the video derives its “story.”

Some of the “T” channels — like the example above — list the source URLs for the images and spoken text underneath the video’s “…More” section. At the same time, many videos do not seem to contain a source list. I’m not sure this kind of data is readily available through YouTube’s API, but I can say it would be difficult to go to nearly 80,000 videos and manually check for lists of sources.

List of URL sources (under collapsed “… More” section) on “A Tease: netanyahu russia jerusalem” video

Everything about these “A Tease…” videos suggests SEO, social politcs amplification, and YouTube AI-playlist placement: In addition to the video titles being keyword-packed and URLs pasted all over the video descriptions, the spoken text that’s already published on news sites should help to boost the overall relevance of these videos and any associated news-related “channels” on YouTube.

TipTop Tech YouTube AI-generated “news” videos are being shared on Facebook

Every Three Minutes a News Video is Born

Given the large number of these AI-generated YouTube videos, I thought it might be interesting to see the time between uploads. It appears some of these “A Tease…” channels (see “A T User” example below) post new “news” videos every three to four minutes.

Video list from API call for “A T User” YouTube channel

Out of the 19 channels that were returned in my initial search, three were different from the 16 “T” branded ones. These channels had very few “A Tease…” videos, but appeared to be part of a larger political video news network. One small Youtube channel named “Trump Vs Putin” (see below), had mostly pro-Trump and “right-wing” titled videos, but also listed several suggesting the videos could be critical and/or pushing specific a political agenda towards Trump.

Sample of videos on YouTube channel “Trump vs. Putin”

FakeTube is here, and AI-generated videos are probably being made faster than they can be identified.

What makes this finding unsettling is that YouTube hasn’t seemed to do anything about this operation. These accounts are prolific, and related ones are being created every few days. Two minute videos aren’t 140-character tweets, so 75,000+ uploads is a lot of chances to draw YouTube’s attention. Especially since I’ve heard about content creators getting into disputes with the company based on its policies related to “keywords” in titles, the use of links on channels, and policies related to copyright.

YouTube network graph of 6,902 AI-generated news videos (Close-up)

How can YouTube not be aware of nearly 80,000 spammy and readily identifiable keyword-packed videos from this network of closely connected “A Tease…” accounts?

Well, I can’t quite understand how YouTube couldn’t know about this operation. Maybe AI-generated video is a new kind of “remixing” — like music (audio) sampling, so it can’t be realisitically enforced? At the same time, current affairs and global politics are not readily comparable to capturing a couple of seconds from another artist’s song; YouTube has clear usage policies and a TOS.

Related to this idea, it’s odd that YouTube’s advanced copyright detection can instantly “tag” obscure songs when you upload videos, and its auto-captioning algorithm can provide “text-on-demand,” but the service can’t yet flag audio read directly from the text of major news stories, or block images that have been pulled from elsewhere on the internet.

Some of the AI-generated “A Tease…” YouTube videos contain the entire spoken text transcript, which are essentially sentences from the sourced news articles, which may further boost the video’s SEO and again increase the chances that it could be flagged on YouTube.

It’s a little worrisome that YouTube either isn’t aware of this huge “fake news” AI-generated video operation, or they’ve just chosen not to do anything about it. I feel this content machine is yet another factor that amplifies the “fake news” problem.

To keep this project’s open data in motion, here’s the first round of data from these AI-generated YouTube channels. This file contains the 6,902 “A Tease…” videos identified in a title search API query for “A Tease: …” And below is the larger dataset with the videos from the channels, if anyone’s interested in doing a more detailed analysis.

I thought I’d add these videos are being pasted all over the internet:

Google searches for YouTube channel IDs

--

--

Jonathan Albright

Professor/researcher. Award-nominated data journalist. Media, data, & tech frmly #columbiajournalism #towcenter #berkmanklein #elonuniversity