Exposing YouTube's Bias!

I have some big questions for YouTube that  they won't answer and that of course made me suspicious, so I have spent the whole two weeks  digging and researching and just trying to uncover the truth for you. So let's get into it. Question  number one is shorts killing long form content on the platform? Shorts the vertical videos that  are less than 60 seconds long were introduced by YouTube back in 2020 prior to that you could  upload a short in length video to the platform but no one was going

to see it because it would  have low watch time so the algorithm just didn't recommend it. Nowadays though if I go onto YouTube  and for example search for 'cake' notice here I'm on the all tab not the shorts Tab and I've got an  ad followed by a row of shorts and after that we have a short another ad, a short and a row of  shorts and another ad followed by more shorts. Surely that has to have an impact on the long form  video views. You've been moving really fast in another area specifically sh

orts? It's definitely  an area we've been investing very heavily in for sure. YouTube has put out plenty of data boasting  about the growth of YouTube shorts on their platform but when I asked them if they could give  me data about the number of long form video views per year since shorts was introduced they said:  we're not able to share any historical data on a platform level on the number of long form versus  shorts views per year. And that's okay it's their platform they don't have to be tra

nsparent with  creators who rely on this for their income if they don't want to. I did find an article in The  Financial Times though where it said that a senior staff member at YouTube had expressed concern that  shorts were cannibalizing long-term views and that fewer long form videos were being uploaded to the  platform and there was a risk of long form dying out. So I think that answers that question but why  does it even matter if it's shorts or long form? Well it matters to creators and to

viewers because  quality content takes time and money to create and shorts is just not paying enough. Zack King got  35 million views on shorts and earned only $343. My next question is related to shorts It's what  even counts as a view on YouTube shorts? I mean historically YouTube has always said that for it  to count as a view on the platform you have to watch for at least 30 seconds of the video and the  viewer has to initiate the view by clicking on it, you can't just autoplay the video. O

bviously  that's not the case with shorts so I asked the question what does count as a view and this was  the reply: we don't have a defined duration for the system to count it as a shorts view but it  does gauge the number of legitimate views for shorts versus those who swiped away immediately.  How can you be counting something that's undefined? That is a really weird answer. So then  I started looking at shorts analytics and you can see that it has the percent who chose to view  versus those

that swiped away. So we know it's not like Tik Tok just swiping past doesn't count  as you've seen the video but if we look at the analytics on this 14 second video you can see here  that on this beginning bit the first 5 Seconds the view is above the 100% line and that's because  once you've seen the video it loops it starts playing again so if you watch it to the end and  start watching it again before you get a chance to swipe away it looks like you are watching more  than once, so depending

where you choose to put your view count at is dramatically going to change  the number of views on the platform. If we look at this longer video it's exactly the same you can  see those first 5 Seconds have a greater than 100% view count. So if you chose to put your view time  at 2 seconds as the defined duration of a short view you're going to have many, many more views  on the platform than if you said it was 10 seconds that counted as a view on a short. So by not  defining it you can manipula

te the number of views on shorts that you're getting on the platform.  The negative of that is by making it all the way at the beginning like 1 second or half a second  is you're going to increase the number of views so the boasting rights on it but you're not going  to increase the CPM. The next question I have is a really curious one. This is one that someone  back at YouTube in Sydney when I used to live there raised with me years ago ... because of all  of the channels that they were managin

g all the female hosted ones had been growing nicely before  the change to the AI algorithm and then by a year after they were all declining rather rapidly.  Certainly in the 'how to' space I noticed a sudden influx of male hosted channels doing really well  and same in the beauty space which those areas both used to be dominated by female hosts and  female creators. But at the same time there was a big influx in the amount of content that was being  promoted for Content Farms which as you know

can contain a lot of misinformation and fake recipes  which is why we've been doing all the debunking on. So I totally forgot about the potential for  a gender bias I just didn't even think about it anymore until recently and then I was watching a  video by Amanda reviewing a conference for YouTube creators that I was considering going to and this  is what she said: this was a dude fest. One maybe two women were on the main stage throughout uh the  3 days the two days you had a lot of uh volunte

ers that were women creators I spoke to a bunch of  them they were great. Now if you're running an event you have to put the biggest creator  creators on the main stage because they're who people want to see and you need more seats in  the auditorium it just makes sense you don't want an empty auditorium with someone smaller on the  stage. So that means either they couldn't get any female creators who were big to come and speak  at their conference or there aren't any female creators who are big

. And that made me wonder and  then I heard Samir say this: our industry is very predominantly male specifically on YouTube. You go  to any Creator event uh you know and it feels like it's it's uh more male-centric than uh than maybe  ever right now on YouTube. Why do you think that is and do you think it's hard for female creators  to break through on YouTube right now? It's a great question. Great question. I'm not quite  sure and it's something I've also been thinking about is is it the natur

e of the algorithm that  rewards more intense content that and maybe I fall in that because I'm doing traditionally  more masculine things sometimes I'm not sure. Some of you might remember back in 2018 Amazon  was in the news for its AI recruiting software. Amazon's software engineers recently uncovered a  big problem. The company has now abandoned an AI recruiting tool after discovering that the program  was biased against women. When it came to choosing an applicant for a technical job like a

software  engineer it was filtering out anyone who had the word women's in their application if they went to  a Women's University or president of the women's chess club or played in the women's hockey team  if any of that was on the resume they were getting lower points and even when they tried to tell  it do not filter based on the word womens we are happy to have women in those roles it was still  filtering out the women. It started looking at the tone of the language used in the application

  the men's ones tended to use words like execute and more aggressive tones I guess and the women's  were more passive tones so it still managed to filter them out somehow so they ended up scrapping  it going well this is not going to work for us because we do want to employ women. The trouble is  AI is always trained on a data set so if there's any bias in that data set it tends to pick that up  and exaggerate it. So because it was trained on 10 years of historical hiring data and over that tim

e  predominantly men had been hired for those roles it had learned that men must be better at those  roles. Bloomberg did something similar where they asked an AI image generator to generate images  of people in different jobs. So they asked for a picture of a CEO a politician a cleaner a doctor  and in all the higher up roles it gave a lot more men than it did women. And you might go well  that's just true it's just showing a reflection of what there is in society because there is a  bias there

still in some of the top roles but the thing is it has exaggerated the bias. For example  if we ask for an image of a doctor it gave nine out of 10 images of a male doctor whereas in the  US 37% of doctors are women. So the bias already existed but the AI has made it even worse. So to  know if there is a bias in the YouTube AI towards male creators we're going to need to analyze a  lot of data and it is hard to access this sort of thing. I'm not interested in shorts views because  they don't pa

y very well so we're just looking at long form views. We can't look at total views for  all time because that's giving us historical data for channels that have been around for a long  time like mine that were around before shorts, so we just want to know how many long form views  channels got in the last 30 days and which ones are at the top. That took hours and days and weeks  of analyzing data and thanks to James my son for helping me with that and doing all nighters and  just crunching numbe

rs but I think we finally have the list of the most viewed channels in the  last 30 days. Before we look at the results I want to highlight another way in which algorithms can  affect our lives as well as deciding which YouTube channels to amplify they also affect how we read  the news. Algorithms will look at your browsing history in order to decide what information  to show you which can really narrow down your perspective. There's an old saying that says in a  court case the first person that

gets up to speak always seems right until they're cross-examined.  In other words you need to hear both sides of a story to get the full picture. Today's sponsor  Grand news is on a mission to help us do just that created by a former NASA engineer, the  website and app gathers related news stories from all around the world into the one place so  that readers can compare coverage. Next to each source you'll see who owns it how reliable their  reporting is and if they have a political bias. Go to

ground.news/howtocook to give it a try. If  you sign up through my link you'll get 40% off the Vantage plan which gives you unlimited access to  all their features, it's a great way to outsmart the algorithms and make sure that you're seeing  the bigger picture. Now back to our results to see if the YouTube AI algorithm is favoring male  creators or not. Imagine that this is the top 100 Channels with the most long form video views  in the last 30 days. These ones are the music channels like Tay

lor Swift. These ones are the  made for kids channels kids watch a lot of content on YouTube. If you didn't know kids creators are  supposed to mark their videos as made for kids just like Bounce Patrol has a great Australian  kids channel if you scroll down you can see there are no comments because comments automatically  get switched off on made-for-kids videos. The other unfortunate thing is your AdSense goes down  by about 10-fold and that's because it's actually illegal to serve targeted ad

s to children. The  reason being is because to serve targeted ads to you they have to track you across the internet  to see what your interests are and what things you are wanting to buy so they can serve those  ads to you and it's illegal to track data on children. YouTube got fined $136 million for doing  so by the FTC so they had to implement a way that creators could say this channel is made for kids  and there will be no targeted ads on that. Anyway back to the top 100 ... these ones are TV

shows  or very large media companies and this may come as a surprise to you but all of these channels are  not in English, which makes sense given that India is the largest consumer of YouTube. So that leaves  us with only five creator channels out of the top 100 so we had to go even further down the list  to get to the top 10 for the month. Coming in at number 10 with 157 million views for the month is  Lankybox it looks very kid orientated for me just so you know kid is defined as under 13 on

YouTube:  do you guys have any idea where the treasure is? Oh they're right it's right there oh behind us oh  you guys are right. But the comments seem to be on for all of the videos so it has not been marked  as made for kids so we will put them on the list. At number nine we have A for Adley. Adley is a  child so I have no idea how this is not marked as made for kids the comments are on for this video  but this one is marked as made for kids. Looking through their latest videos it seems to be

50/50  whether they've marked it as made for kids or not so because half of them are marked as made for  kids I think we'll just put their channel in the made for kids section and take it off this list.  Which means we need to go back one further down to number 11 which will now be number 10 which  is penguinz0 ... all right let's ride boys. He does streaming and commentary videos and got 154  million long form views in the last month. It's crazy. Add him to the list and at number eight  we hav

e Better Bocuk this channel also has a kid as the main character but comments seem to be on  for every single video that I looked at so it's not marked as made for kids so we'll need to put  it on the list. Next at number seven is Maizen this one's run by a couple of guys it was the top  gaming channel in Japan and they now also dub it into English. It's mainly Minecraft and it looks  kiddy to me again but the comments are on so we'll put them on the list. Next at number six is Dhar  Mann Studio

s with 180 million views the channel is run by Dhar man and he appears in the videos. I  don't even like Dhar man anyway what Darman I was just kidding. We don't have any females yet.  Number five is Nile Red it's great to have a science channel doing so well it just seemed weird  to me that something so toxic could smell good. This channel is run and hosted by Nigel so we'll  add him to the list the list is full so we're going to have to shrink down the guys so that we  have some more space the

re. This next Channel at number four had 202 million long form views last  month. The channel has animations created by Alexi Garasimov and his latest War series Skibbidy  Toilet has really taken off let's put him on the list well actually he doesn't actually appear  in his videos so let's put one of his toilet characters on the list for him instead. Now next  at number three we have Aphmau with 257 million views again it's Minecraft and looks kiddy to me  we're going to play a game but the comm

ents are on. One thing for parents to be aware of is that  some kiddy looking channels have added more adult conversations or adult themes to some of their  videos in order not to be marked as made for kids and that seems to be the case with some of the  content on this channel she says on Tumblr that not all of her videos are suitable for kids to  view. So just be aware of listening to what your kids are watching not just them with headphones  and you can only see it cuz it's not necessarily go

ing to be appropriate but we do finally have a  female on the list so that is exciting. Next at number two we have JoJo Sim with 260 million views  for the month. Is it a lie or is it the truth that a Subway foot long is actually a foot long. So  another guy and at number one with 830 million long form views for the month is of course  Mr Beast. You have exactly 24 hours to build whatever you want to protect your Lamborghini from  the bullets. The timer starts now. So if we add add him to the to

p of the list that makes it nine  male creators to one female creator or 2.3 billion video views compared to 257 million video views.  Which seems pretty biased but if we shrink that down and add the next 20 channels on the list we  end up with 20 male creators seven female creators and three couples or family Vlogs so I put them  in the middle because they've got both, so that seems a little little fairer still biased but a  little fairer but that was until I looked at those seven female channe

ls. Of those seven channels  six of them the female doesn't actually appear in the videos at all. Three of them are gaming so  they appear only as an animated character one does videos about other people in their dogs and they  are not seen personally in the videos at all one does ASMR slime so they are not seen or heard in  their videos and another has a monkey which surely this belongs in made for kids but comments are on  so I put it on the list. So the question could be asked does AI even kn

ow that those channels  are run by a female? Certainly historically before AI YouTube always used to have a male  at the top of the Creator list there was Smosh, then Fred. then Ryan Higa, followed by Ray William  Johnson, Jenna Marbles was doing really well but she never actually made it to the top spot before  PewDiePie took over and now of of course there's Mr Beast. The question is has AI looked at that  historical data and then exaggerated a bias towards male creators? I really don't know b

ecause  there are millions of YouTube channels and we have only looked at the tip of the iceberg certainly  what we can see doesn't look great. Let me know what you think in the comments out of interest  I asked mid journey and AI image generator to generate me some pictures of Youtubers  🙋🏻‍♂️🙋🏻‍♂️🙋🏻‍♂️🙋🏻‍♂️🙋🏻‍♂️🙋🏻‍♂️🙋🏻‍♂️ With thanks to my amazing patrons for all of your support of  this channel that allows us to keep going despite YouTube algorithms. Make it a great week by  bei

ng kind to others and I'll see you on Friday.