Your VK video views look lower because VK changed what counts as a view, and the change arrived at the start of 2026. On a long video, a view is now recorded only after five seconds of watch time. It used to register almost as soon as the player loaded, which meant a thumb scrolling past your thumbnail was worth the same as somebody who genuinely watched. The same audience behaving in exactly the same way therefore returns a far smaller number today. Your reach has not collapsed, and every benchmark you set for yourself before the change is measuring something that no longer exists.
Table of Contents
What the five second rule counts and what it quietly deletes
The old counter was generous to the point of being useless. Autoplay in a feed, an accidental tap, a click straight back out, a tab left open in a browser, all of it counted. The new threshold brings long video in line with how the rest of the industry counts a view, and it does one thing very deliberately. It deletes the portion of your old numbers that was never watching in the first place.
Work out what that means for your own archive. If a third of your old views were feed autoplays lasting two seconds, a third of your history evaporated on paper without a single human being changing their behaviour. The drop is not even across your catalogue either. A video that opens on a static title card loses far more than one that opens on movement, because the five second gate sits precisely where a slow opening does its damage. Two videos that used to look equally popular can now report wildly different numbers.
Clips and long uploads are not counted by the same rule
The five second threshold is described for long video. If you publish both short clips and full length uploads, stop pooling them in one spreadsheet, because you are averaging two different definitions and the result means nothing. Split them. Compare clips to clips and long videos to long videos, and do it from the change onwards.
The May 2026 recommendations change moved your reach around too
The second change is easier to miss because it never shows up as a number on a dashboard. In May 2026 VK reworked how recommendations distribute content. The new behaviour pushes a video out to a broad audience straight away instead of waiting for a creator to build a subscriber base first, and the system reads the topic and the author profile in real time, then predicts the response before deciding where it goes.
For a brand new account that is genuinely good news. For an established community it removes an advantage you had been quietly living on. Your subscriber count used to guarantee a floor of distribution on every upload. Each video is now assessed much more on its own merits, so a weak upload from a large community can travel less far than a strong one from a small community. If your view counts fell while your subscriber count held steady, this is the second half of the explanation.
Weak retention now drags down everything else you post
Here is the part to take seriously, because it changes what is safe to do.
Poor performance signals no longer stay contained inside the video that produced them. They accumulate at the level of the community, and the recommendation system becomes more pessimistic about everything that community publishes. One badly opening upload will not sink you. A run of them, posted quickly because you were testing, can pull organic reach down across the whole page including the posts that were doing perfectly well on their own.
So experiment with more care than you used to. Space your tests out. Keep publishing the format you already know works in between them, and give each test a fair run before you judge it. If you want to try something genuinely unfamiliar, consider testing it somewhere it cannot cost you anything first, then bringing only the winner to VK.
The numbers that still compare across the change
View counts from before and after belong to two different measuring systems, and putting them on the same chart produces a lie. These survive the change intact.
| What you were tracking | Why it broke | Track this instead |
|---|---|---|
| Views per video | The threshold moved, so old and new figures are not the same unit | Retention past five seconds as a share of players started |
| Views measured against subscriber count | Distribution no longer leans on subscribers the way it did | Average watch time and completion rate |
| Total community views per month | Mixes both counting systems inside one figure | Community reach and unique viewers |
| Best performing video of all time | Anything from before 2026 is inflated relative to now | Best video since the change, ranked on watch time |
| View to subscriber conversion | The denominator quietly changed meaning | New subscribers per video, counted directly |
Retention is the one to build your habits around. It was already the honest number and now it is also the stable one, because a percentage of people who kept watching does not care where the counting threshold happens to sit.
Rebuild your baseline from the change date
- Draw a line at the point your numbers stepped down, and treat everything before it as a closed archive.
- Collect at least eight to ten videos published after that line before you decide what normal looks like now.
- Record retention, average watch time and completion for each one, plus the subscribers it brought in.
- Set new targets from the median of that group rather than from your best month of last year.
- Review monthly, and expect the figures to keep drifting while VK carries on tuning recommendations.
If you report to a client or a manager, do this before anybody asks. A chart that steps down at the start of the year with no explanation attached invites the wrong conclusion, and the wrong conclusion is usually that you stopped being any good at your job.
The opening seconds are now doing the counting
Every guide on the internet tells you to hook the viewer in the first three seconds, and you were expecting me to repeat it here. The number moved, and it matters more than it did before. Five seconds is where the counter now sits, so a five second logo animation means the view is never recorded at all, however good the following ten minutes turn out to be. Your channel branding, the slow fade in, the sponsor card, the cheerful welcome back greeting, all of them now cost you the view itself.
Practical version. Open on the most interesting frame you own. Put visible motion in the first second. Get whatever the video promised on screen before second five, and say the payoff out loud while you are at it. If you have an intro sequence you are fond of, move it to a point where people are already committed, or let it go.
What to do with the drop itself
A number that fell because the measuring rule changed says nothing about the quality of your work. All it records is that VK swapped the ruler for a shorter one. The audience that watched you last year is still there in the retention curve, still finishing your videos at the same rate. Reset your baseline, guard the first five seconds like they are the whole video, and stop comparing today against a counter that VK retired.
