A YouTube ‘view’ does not necessarily mean a viewer who has actually chosen to watch a video. And since 24 August 2026, this has become even more apparent: YouTube has changed its public definition of a view. From now on, for Shorts as well as long-form videos, podcasts and live streams, a view can be counted as soon as the video starts playing, from the very first frame. This includes, in particular, certain auto-plays on the homepage. At the same time, YouTube maintains a separate metric in Analytics, the ‘engaged view’, which more accurately reflects someone who has chosen to watch or has continued watching beyond the first few seconds.
This already significantly changes the meaning of the large counter displayed beneath a video. 10,000 views do not mean 10,000 people have watched the video. Essentially, they mean that the video has started playing 10,000 times according to YouTube’s counting rules. Some people may have watched right through to the end credits, whilst others may have watched for just a few seconds. The figure is real in the sense that YouTube has recorded it, but it only tells part of the story.
And then there are the genuine fake views.
For years, there has been a black market where you can buy views, subscribers, likes and comments. It takes very little research to come across sellers promising thousands of views. Behind this may lie automated systems, networks of accounts, incentivised traffic or various techniques designed to artificially give the impression that a piece of content is resonating with its audience.
YouTube explicitly prohibits these practices. Its rules forbid the use of automated systems, ‘viewbotting’, the purchase of artificial traffic and mechanisms designed to artificially inflate views, likes, comments or subscribers. A channel may face penalties if such practices are detected.
But this is where the story gets much more interesting: YouTube does not necessarily detect all fake views straight away.
A scientific study published in 2024 in *Scientific Reports* examined around a thousand French channels over a period of one and a half years. The researchers observed the adjustments made by YouTube when it removes views deemed to be artificial or illegitimate. More than 78 per cent of the videos in their dataset had had their view counts adjusted. The researchers also found that these adjustments could occur in waves and relatively late in a video’s lifespan.
This is a crucial detail.
Imagine a video that suddenly receives 50,000 artificial views. For a while, its view count may give the impression of success. People may click on it precisely because it appears popular. And the hypothesis examined by the researchers is even more troubling: this artificial popularity could potentially contribute indirectly to generating real popularity, particularly through recommendations.
Care must be taken, however, not to treat the hypothesis as a certainty. The authors themselves emphasise that they cannot definitively establish that fake views subsequently lead to more recommendations. They observe a correlation and explain that they lack internal data held by YouTube to demonstrate causality.
This is precisely where YouTube’s real problem lies: its lack of transparency.
We constantly talk about ‘the algorithm’, as if there were some secret formula somewhere that decides: this video deserves 100,000 views, that one 300. In reality, YouTube describes an extremely complex and personalised recommendation system. Viewing history, searches, subscriptions, likes, dislikes, content flagged as ‘not interesting’, and the behaviour of people with similar habits: the platform claims to use over 80 billion signals to fuel its recommendations.
Two people opening YouTube at exactly the same moment therefore almost never see the same YouTube.
And above all, success is not proportional to the intrinsic quality of a piece of work. A remarkable video may remain unseen, whilst a mediocre one goes viral. The algorithm does not have a little film critic tasked with determining what is good. It essentially tries to predict what a given user is likely to watch and enjoy.
This is why the start of a video is so crucial. YouTube constantly pits it against fierce competition: according to its own explanations, a video is not only compared to other videos on its channel, but also to other content that might interest the viewer. General interest in the subject, competition and variations in audience behaviour also influence its reach.
Let’s now add a third, much less well-known category of views: paid views, which are perfectly legal.
A content creator, a record label, a company or a distributor can purchase an advertising campaign to promote a video. This is not fraud. YouTube actually identifies a traffic source called ‘YouTube advertising’ in Analytics. Depending on the advert format and viewing conditions, certain ad placements may be counted as views.
We therefore have a much more complex picture than simply ‘one view = one interested person’.
A video may receive organic views from people searching for it; views from recommendations; views from external sites; automatic plays; perfectly legal advertising traffic; and artificial traffic that will be detected. Artificial traffic that isn’t immediately identified as such. And behind all of this lies another, far more interesting piece of data: how many people actually chose to stay?
This is where creators should probably start looking at YouTube differently.
The public view counter is spectacular because it’s visible. But to truly understand a video, Analytics data is infinitely more informative: average watch time, audience retention, traffic sources, unique viewers, impressions, click-through rates and, now, the distinction between views and engaged views. Since August 2026, YouTube has, in fact, explicitly recognised this difference: revenue from the Partner Programme continues to be based on ‘engaged views’ and engaged hours, whilst eligibility criteria are based on qualified views.
In other words, YouTube itself does not assign the same economic value to all ‘views’ displayed publicly.
And this is perhaps the best way to understand the paradox.
YouTube’s figures are not false. To say that ‘almost everything is false’ might be a tempting phrase, but it is factually exaggerated. What is false is the intuitive meaning we attribute to them.
A million views no longer means ‘a million people have watched this video’. One hundred thousand subscribers do not guarantee one hundred thousand viewers. A video with 500 views is not necessarily an artistic failure. A video with two million views has not necessarily gone viral spontaneously. And two videos with exactly 50,000 views each may have radically different audience realities behind them.
YouTube does indeed combat fraudulent traffic. The platform states that it analyses visits to determine whether they are valid and may remove traffic deemed to be artificial. It also acknowledges that certain services claiming to help grow a channel may send artificial traffic, sometimes even without the creator’s knowledge.
The fundamental problem, therefore, is less the existence of a massive conspiracy fabricating all the view counts than the public’s inability to know exactly what they are looking at when they look at a view count.
And this nuance is staggering.
YouTube has got us used to measuring culture in the same way we measure temperature: 412 views, cold; 70,000, warm; 10 million, red-hot. Artists, journalists, producers and sometimes the creators themselves end up confusing visibility, audience, attention, influence and quality.
Yet these are five different things.
YouTube’s greatest magic trick may, in the end, not be its algorithm. It is having succeeded in condensing human behaviours—which sometimes have almost nothing to do with one another—into a single figure: ‘views’.
And yet we continue to regard this figure as if it told the whole truth.
