Ex-YouTube Employee Explains “Shadowbanning”

What is YouTube analyzing when deciding whether or not to show their videos to their users? YouTube has this multi-layered system of reviews. True or false, Shadowbanning is real. Hi, I'm Alina Verbenchuk. I'm a former YouTube product manager and recently I published a book called "How Platforms Work" and it's based on my experience working at YouTube. Okay, quick disclaimer, this is a pretty intricate episode. We're speaking with someone who actually worked on the algorithm at YouTube and so we get really detailed about the nuances that take place there. Alina, I am really excited to speak with you today because you know the back end of YouTube, the mysterious YouTube algorithm and YouTube studio and all those kinds of crazy things. Let's say I just uploaded a video to the internet and everything went great, but I got the yellow icon of the video getting flagged. What led to that and what are the next steps if I want to challenge that? When you upload a video to YouTube, it kind of gets canned across policies like community guidelines, mentalization policies and copyright simultaneously. So if you have something that infringes any of those three things, you will see it after you upload your content and you're either going to get, I don't know, your video might get blocked because you violated community guidelines, which is probably the most severe one, or you might get demonetized, like not suitable for all audiences to be demonetized for all advertisers. What goes into the analysis of the video during its upload that might cause it? I assume the transcript is helpful for finding inappropriate language, but also is it scanning the video pixel by pixel and making assessments of whether something is safe or not safe? All information that you upload to YouTube goes into the decision, into the verdict. So, A, your video file and video file has images that literally, this is like a fingerprint, they're like basically your video is like cut in, I don't know, millions of different images and machine learning, basically the machine analyzes what is in the video? Is it something good? Is it something bad? Is it something, I don't know, inappropriate, harmful actions, violence or it's absolutely normal? Then it's audio, you're absolutely right, so an audio might, there might be a lot of, I don't know, inappropriate language or I might encourage someone for harmful actions, all of that also goes into into information then and it's not just about the file itself, it's also the metadata, the thumbnail, the title, the description, everything goes into information, like into basically into the verdict about whether the video should be shown or not. There are some implications on how these things and how metadata influences the recommendations because some people are like oh, description is not important or the title is the main, like spoiler alerts, the title is the key thing and the thumbnail is very important, but everything else still supports your, basically how you're going to be recommended and if you, the more information you give to the algorithm, kind of the broader it can be recommended so the better so basically you can add more information to your description to give algorithms more information about who this video might be suitable for. This is a really interesting thread that I want to kind of go down, but I do want to come back to copyright in just a moment, but let's, let's go into that further. So you mentioned you said spoiler alert, title is the most important, thumbnail is also exceptionally important, I'm hearing description is important, like everything is important because you're trying to give it the best understanding of what content you're actually making. A lot of advice out there is saying don't pay attention to tagging content, hashtags have kind of lost their energy, are you saying it should be everything? You should go all in on all components, fill it out as best as possible. Where does the hierarchy go beyond the packaging? I would say don't over index on description, so description is like a support material, support material, it's not the most important thing. Basically try to put as much effort and as much attention to the title and to the thumbnail. The reason for that is that these are the things that are visible when people are going to click your video, and basically the more people click your video, the the higher the CTR, the better, but description is like supports it, so basically if I had, I don't know, one hour to spend on all metadata, I would spend probably, I don't know, 50 minutes on title and thumbnails on the thumbnail, and I don't know, 10 minutes on making sure the description kind of shows everything, reflects everything that I have in my video. But that's from like a creative strategy perspective, I think. Like I think you are talking about it from, you know, how to get more clicks or like what you should prioritize to make the audience more engaged with your content. I'm curious about from like an algorithmic back end perspective, like what is YouTube analyzing when deciding whether or not to show their videos to their users. That's exactly the kind of the misconception creators have that's basically the algorithm and creativity that these are two separate things. No, it's not because the algorithm essentially is, it reflects the audience interest. So whatever audience is interested in, whatever they're like, they believe they want to watch and they want to click on, this is something that you should optimize because description is not visible immediately. You might spend a bit less time and effort on optimizing the description. It still might be shown to people kind of like to look alike audiences. Let's let's call them that way, basically to similar audiences that would watch a video. But the strongest predictor of how the video is going to be recommended is still your title, thumbnail, and basically everything that is visible on the surface before a viewer clicks your video. So basically creativity and algorithm, it's essentially the same thing. And people think that, oh, there's this evil algorithm that doesn't recommend my video. No, it's actually people that are not interested for some reason for various reasons. Yeah, I want to go on the record saying that I am not anti algorithm. I don't think the algorithms evil. I regularly tell people that like this home that I live in right now was purchased thanks to the great work of people like yourself on YouTube creating a job for someone like me. So I'm very grateful. And I think the algorithm is awesome. And I think this platform is awesome. I want to make sure that's clear. I want to double back to the videos getting flag. This is what we just talked about. It has been decided that my video has something that goes against YouTube's policy. What if I disagree? I'm going to challenge that. What happens on YouTube side of things? So yeah, it's a great question. And I think another myth is that if I appeal the decision, nothing's going to happen. But it actually something might happen because YouTube might change the decision. And usually the first step of any review when you upload a video, this is the machine review. And if machine flags something, it goes to the human review. Like in simple words, let's just simplify it. And then as soon as you let's say, I don't know, you got some kind of a restriction on the video, let it be let it be monetization or community guidelines or even corporate, you can appeal and you can dispute the decision. And with that, it might usually go to the human review unless it's like black and white and you clearly use some kind of like policies that are very strict and clearly you basically it's seen from technical perspective that you're like circumventing the systems. I think that's unlikely. Like we obviously know what is a breach of policy. But when it's uncertain, that's where it gets a little bit weird. When the creator says, what are you talking about? I didn't do anything rude or incorrect or you know, nothing illegal is being shown on camera. That's what I'm curious about. So you mentioned that it goes to a human review. And is it a single person making a single decision? Is it a group of people coming together and scratching their heads and saying, actually, no, I think this is okay? Well, I can't give you all the details of how it works. Also, processes change. But to simplify it, it's always a mix. Basically, it's a mix of kind of human review, of machine review and human review. It might be clear. It might require some discussion in the team. It's case by case basis. This is why people sometimes like creators, sometimes are like, why don't you clearly state it in the help center? What's allowed? What's not allowed? It's just impossible because every case is so different and is very unique. And sometimes, as you mentioned, it might require like a team of people coming together and making a decision. Just one recommendation that I have is the more explanation you give, the better, because how creators usually behave. They're like, right, and oh, I'm not happy with the with the decision. Replease review. No context at all. So if you give them more context, especially about, I don't know, actually about everything. Let it be copyright or community guidelines. With copyright, they might require some legal documents. And very often, creators just don't have any kind of legal support. And they're like, oh, I'm right. You might be right, but like, can you prove it? True or false, shadow banning is real. False people perceive so many different things for shadow banning. And they assume that they were shadow banned for basically wrong reasons. And there was no shadow banning. And shadow banning does not exist in the way that people think of shadow banning that say there is a YouTube or Instagram or six or complete, there's like, oh, I don't like this account. Let's just shadow ban them. Is that what you mean? Yep. people feeling suppressed and particularly over multiple videos as well. Maybe someone has had recent success for five or six videos and then all of a sudden nothing. What I've seen in my practice as a creator, because when I was a tutor, like I couldn't experiment that much, but when I became a creator, started to experiment, when I had this thought that, basically, I saw some video, for example, I was posting consistently and then I stopped posting for like a couple of weeks. And then I started posting again and my account was kind of shadow-bound because I didn't get as many views as they expected. And I was like, "Oh, this is probably how people feel and how they think of social media that this is shadow-banning." But it's not shadow-banning, it's just like I was absent for two weeks. So here's like one reason, like I lost consistency in my posting. Another example, I was talking about topic X and then I suddenly wanted to start to talk about the topic Y. And this topic Y isn't interesting to my audience, like nobody cares about my audience. And I also see the decline immediately. Or there is seasonality, or there is, again, change in my format. There are so many things that are mistaken for shadow-banning. This is why shadow-banning doesn't really exist. There are things that can limit your, to kind of to your point about shadow-banning, but what can be perceived as it, and it was real, are limitations and restrictions on the video. And they might happen for policy reasons, for, let's say, age restrictions, mentalization restrictions. But again, mentalization doesn't restrict your reach. Mentalization is usually, there is a correlation between you violating, or like you being, you making borderline content. And this content can't be shown to, let's say, underage audience. Which means that people who are logged out of YouTube, they won't see this content. So YouTube doesn't show sensitive content when it doesn't, it's not sure in the age of the audience. It's funny because you mentioned some characteristics that I would relate to human behavior. Like if you post consistently for a week, every single week, for a year on end, and then you take a two-month break, and then you come back and you feel shadow-baned, I would suspect that's actually your audience not having, like, the repetition of seeing you pop onto their feet every Tuesday, and so they've gone and they've done other things. Yeah, and I'll give an example from, if I can give an example from Instagram, because I think in on Instagram, people feel it even more than on YouTube. I did this research for my book, and like Adam Maseri, or I think someone else, in their help center, they literally wrote it out that by 2016, or something like that, like basically very early on in the history of Instagram, 50% of the audience, or like 60% of the audience, hasn't seen all the posts that they friends were posting, because it's already been too noisy, it's already, it's just too noisy. And imagine like we are 2026. Imagine how much content is produced and uploaded every second on every single platform. So you can't really say that all it's the platform to blame is, again, people have their lies, and people, you're competing with so many creators and so many things, there's TikTok, there's YouTube, there's Netflix, there's this, there's that, there are like, I don't know HBO. There are so many entertainment sources that people are just like, it's not always gonna happen. That's a great point too. I think from like a creative strategy perspective, personally in my experience using social media, I am not as personally attached to creators as I once was, like 10 years ago. I used to follow creators every single video that they made, no matter what the topic was. I was so excited because they were part of my life. Now I'm much more interested in like topics, and if a creator that I like touches on that topic, I'm gonna go and watch that video, but if they touch on things that I'm not particularly interested in, there's so much out there now than there was 10 years ago that I'm not personally attached to people anymore. So my behavior has changed, and I suspect that's the case with many people. Yeah, and I think if I can add on that, like, I think short form videos changed, changed the game for everyone forever. In terms of, as you mentioned, like, following a creator versus following the topic. I don't really see this trend anymore that people are chasing kind of subscriber numbers, followers numbers. I mean, they correlate, there is a strong correlation we can't deny it. But this whole thing about the reach, that one, let's say short or real, or tick-tock, can get viral, and you might actually get more out of like one piece of content. You're now thinking as a strategist, you're more thinking of assets rather than, okay, channel. And this channel is, yes, it is real estate, real estate, but like your asset is an asset, and one asset might be, might get so big. Yeah, we literally call videos assets now rather than videos. And it's a little soul-sucking because I like the old, pure version of YouTube that I grew up on. But the reality is, like, we think of things as assets now, and we definitely try to find ways to take a long form video and turn into a short form video and turn into a carousel post on Instagram. Like, we're constantly, you know, repurposing that content. So it is more of an asset than a single video. One of the things that I've been struggling with with YouTube in particular has been, for the past 20 years, the value proposition from about YouTube that's different from Instagram, that's different from tick-tock, is it's a click-to-watch platform, and yeah, as a creator, you get people who spend six minutes, eight minutes, 20 minutes, an hour of their time with you, and you get a deeper relationship. It was a little disheartening, but I understand, but it was a little disheartening to see YouTube go the short form route and kind of copy Instagram and tick-tock and try and take up some of that. There's been some conversation online about YouTube steering further into shorts, and like even removing some of the long form options on the browse page, like there used to be eight tiles to click from, eight videos to click from, now there's like three. Now there's a bunch of shorts, now there's a bunch of games as well, which is crazy to me on YouTube. How do you think about the value prop of YouTube compared to other platforms? One thing before I start answering the question, I'll mention that the homepage is personalized. So what you see and what other people see, my Diffa, and there are always some kind of experiments going on. What I mean by that is that let's say shorts, shelf, goes up, goes down, like the tiles, all that stuff might change. It's normal, so it means that teams might be running some kind of experiments to see what is more interesting for the audience or less interesting for the audience. But in terms of the kind of strategy wise, it's hard to kind of blame YouTube for going short form because also human behavior changed quite a lot. And you didn't mention it, but I will mention it like what also came into question. So we had desktop that we had mobile, and at least like 60% of the traffic came from mobile. This is short form, vertical format is easier to watch a mobile and kind of not necessarily obviously for kind of competitive reasons YouTube wanted to go that way. But from the even from human behavior reasons, it also went when there so from user experience perspective. What YouTube I don't think they counted for and what changed in COVID was TV, and now also YouTube goes into TV quite strongly and long form. So kind of coming back to the value proposition of YouTube is that you can find any type of content included the content for your living room for your TV devices and YouTube is kind of like 360. You can watch it from your mobile phone, from your desktop, from TV as well. And it can replace like any type of entertainment for you. But in terms of the, I think the ecosystem, how the ecosystem is built is that it still has the most advanced monetization mechanism or like monetization ecosystem for creators. It still pays more than any other platform. It's still the fact that it has long long form content. It's great both for creators in terms of the monetization because you extract much more from long form videos. It makes a little sense for brands who want to advertise because being remembered on 30 second video is so much harder than being remembered from like one hour video and all these platforms exist because of advertisers. So while advertisers will be choosing longer format, long format will exist, which means better value for creators as well. Yeah, I do agree that YouTube is the best place to monetize natively on the platform. I do think that it's changed in the past five years. And one thing that I'm realizing through this conversation is like I have so much affinity for what YouTube was. And even from the monetization perspective, it was so easy, like it's not easy, but you're making so much more money than what's available now. But I don't think that's because of YouTube. I just think that's because marketing teams have now gotten much smarter and the surface of the creator economy has gotten much more broad. And so there's more people to cover and marketing teams actually have budgets allocated. So there's less money to be made as a whole for per creator. That's just something I was thinking about as you were talking about monetization. I do think as a whole YouTube is the best place for native monetization. You talked about TV and I saw that YouTube is now assessing whether you're watching TV alone or with like three or four people. How are they doing that? That's crazy to me. I mean, it's it's love that they're spying on you. I would say it's more of a calculation that goes into and again, it's more on the basically advertisers will feel this more than creators. Creators might feel it's slightly when they see and again, like a uplift in their revenues when YouTube decides to kind of charge the advertisers slightly more. And it's a bit of a like a coefficient or like multiplayer that's YouTube is figuring out and I basically I can't provide more details. whatever is like written on help centers, et cetera. I think it's the fullest information, but it's math. It's not like someone is spying on you on like, okay, are you watching it? It's the two or two people or three people or four people. - I want to talk about YouTube Studio for a moment. You worked on the back end of YouTube Studio, built out some of the data that's presented there, I assume. It's gotten a pretty significant upgrade in the last year in the Ask feature. - Everybody has their own YouTube strategist. Is it a fantastic tool? Is there anything, is there any way that you're using YouTube native AI that could give people the upper edge? - So I was there when kind of the feature the idea was just born, and it was long, long before any AI tools before Chan Jipiti, like two years before Chan Jipiti, like long time ago. In terms of how do I use that feature, what I can tell you for like for sure, for like 100% sure. And again, like I respect all the players in the market. There are various tools that are doing similar things, but the thing with kind of native YouTube tools is that they are built on internal data and this data is first-party data. So it's not built on any kind of scrapping API, assumptions, predictions, machine learning models because there are various platforms and players in the market and companies, and big companies that kind of build in it as a prediction. So with YouTube, you can be 100% sure that it's built on your data, it's first-party data, it's real, and whatever, it tells you it's very close or like 99% true to what your audience signals to the platform. - Is there any other place in YouTube studio that's more underrated than that tool? - I think the most underrated place, I'm biased. Sorry, I was working on YouTube analytics for such a long time. YouTube analytics is the most underrated place in the whole YouTube studio. I think it can tell you so much more than any kind of like AI research tool or any other content planning to let it be internal, external, whatever. What I learned being like working with YouTube analytics is that YouTube analytics is a bit like, I would say like a performance review to your work that is like happening every single minute, it's live 24/7, it kind of reviews your work. It's not a comfortable place to look at. It's like as if your boss was there 24/7 overlooking your work as a creative strategist, right? And giving you feedback every single minute. So emotionally, I think it's a hard place to look at because it's not always what you expect. But then this is the place that gives you truth whether it's like hard truth or good feedback that people are enjoying the content that you watch. - It's like a drug a little bit 'cause when you get a win, it feels so good. And then when you get a loss, you're like, no, that's not what I wanted. - That's what I mean, it's underrated for a reason. It's a bit stressful. So you're like, I don't wanna look there. - But you should, actually I find the advanced panel of YouTube analytics to be kind of overwhelming sometimes. And I have caught myself thinking I should spend more time in there, but not because it's pretty daunting. That's what I'm saying. It is daunting. It is a bit intimidating, but it's really hard to make it less daunting because there are people basically, if you think about, I don't know, 80, 20 or all, 80% of people wouldn't even look there. They just like, we're happy with like basic cards and metrics. We don't wanna look in the advanced mode. In 20% of data enthusiasts and people who are really into being very kind of analytical data driven, they will be there and being like digging into the daytime, being a bit obsessed about optimizing the metrics. And I know people like that and they still, sometimes they approach me and be like, hey, that this metric differs for like one digit is like something changed like. And like you can just relax because it won't, I don't know, change the whole game for you. Like yes, if it was, I don't know, 15, 20, 30% change, that's a difference, but when it's just like, I don't know, couple of percent doesn't influence anything. - What's tricky to me is it presents all this data, but then the question after you look at this data, it's like, well, now what? Like I don't always know how to influence the data. I remember one instance where I was trying to improve our AVP, our average views per viewer. And it was always stuck at like 1.2 and then we would upload another video and be like 1.2 and we try another tactic, we 1.2. And it's just like, I don't know how to influence this any better. It's one thing to have the data, but then also YouTube isn't necessarily recommending like, oh, here's how you improve this one particular data point. And that is also kind of daunting to me. - I totally get it. But I also think, unfortunately, unfortunately, kind of it's not necessarily YouTube's job to tell you how to improve your content and like on the creative side of things. And this is where kind of human creativity takes play, like comes into the play and you should probably think of it yourself strategically. Also, every channel is different. Another reason why it's hard for YouTube to do it is that every single channel, every single creator is different, having, they have different goals. And they might optimize different metrics. So what's good for one channel might be absolutely disastrous for the other channel. And people are doing different things. Some people are chasing reach, some people are chasing monetization, some other people, like, I don't know, there is a, let's say, there is a cultural channel and they're like, oh, non-profit channel. And they're like, we're not monetizing at all. - It's kind of crazy. Like I started watching YouTube when I was much younger. I think it was like 2010, 2011. And I was really creative and I like taking pictures and making movies and all that kind of stuff. And YouTube felt like a great place for that. And then it's kind of gotten into this, like that naive, pure movie making teenager. Everything has changed and everybody's gotten involved. And now it's trying to qualify creativity, which is inherently like subjective. But the data is making it objective. It just kind of fries my brain. - Yeah, I hope it's not too kind of daunting or intimidating because obviously part of it is just being creative and having fun or like enjoying the process with the content. Because as content creators, again, it depends. If you are doing, let's say, you're managing a business channel or a brand channel. Yes, you treat everything as an asset. You treat everything as a business metric. You treat everything kind of extremely data driven. But if you're, let's say an entertainment channel and you're a creator, you're yourself on camera, you're engaging with your audience and you want to build relationships with them. Very often the content that you predict or you polish or like you invest time and you like, you analyze everything, average iterations, I don't know, dips, spikes, all the stuff. Okay, I'm doing these best practices that work for me, not doing this stuff that did work. And then that video flops and you're like, what did I do wrong? And this is like part of kind of us being human will also remember that they're also humans on the other side. They're also people and they might be interested in something and might be completely ignoring some other thing. And I still believe as much as I want to be like data driven kind of authenticity rules, if we're talking about like kind of personal channels or personal branding channels. It's not black and white, it's like art and science. So I always think of it as art and science and with a big, big part of arts in it, not just analytics. I always wrap up my interviews by asking my guests if there is any creator out there that is really exciting to them. Someone who is on their radar that's worthy of a shout out. I'm really into, because I'm also building my company and I'm into like business content. I personally love and I enjoy watching Cody Sanchez. The business advice and all the stuff that she says is very relevant to me. Nice. Well, I did work with Cody Sanchez for a number of months and I was a strategist for her. And every time I work with someone, any thought leader, anybody, I kind of like try and boil down who they are into a single sentence or into a single lesson. And with Cody, it was speed. She works so quickly. She ships stuff at 80% good and then adjusts on the fly. That was the one takeaway from her. And I found her so fascinating to watch because she's juggling 100 different businesses. She's a YouTube creator and Instagram influencer. She's so many different things all at once. And the only reason that she can do it is because she acts so, so quickly. She put a video out like two or three days ago that completely outlines her social media strategy. I just found myself nodding to the entire thing. Like what she said in that video is completely accurate to the way that she and I communicated and she communicates with her team. It's a fantastic video. Thank you so much for watching or listening to this episode. Alina actually wrote a book called How Platforms Work. And I read it and found it super enlightening. So if you enjoyed this episode, that's probably the next place you wanna go. Buy her book, it's excellent. If you're new to this channel, please consider subscribing or leaving us a review on the audio platforms. We also interviewed a colleague of Alina who also worked behind the scenes at YouTube. And it was a fantastic interview. She got us over 100,000 views on YouTube, which is crazy. Her name is Nastya and you can check out that episode here. We will see you next Thursday. Thank you so much. See ya.