The wrong audience is worse than none

Every recommender builds a picture of who a piece of content suits, and it builds it out of who responded to it. That picture is what decides who sees the next video.

So attention has two effects, not one. There is the immediate effect — a number goes up — and there is the durable one: the system updates its model of your audience. The second matters far more and almost nobody discusses it.

Suppose a hundred people who care nothing for your subject watch your video and move on. Watch time is short. Nobody comments, nobody saves, nobody visits your profile.

The system did exactly what it should: it showed the video to people, and they were not interested. What it concludes is that the video does not hold attention — the same conclusion it would draw if the content were genuinely poor.

Now it has a picture of your audience assembled from people who were never your audience, and it uses that picture to choose who sees the next one. You have not been penalised. You have been accurately delivered to the wrong people, and taught the system to keep doing it.

This is why services selling a fixed number of views to a video do not work even when nobody is caught. The mechanism that defeats them is the recommender working correctly.

It is, and not by inspecting individual actions. Coordinated attention is detectable by its shape: forty identical actions in four minutes, saves arriving with no watch time behind them, the same two accounts trading every day, a video whose saves match its views one for one. None of the individual actions is fake. The distribution is.

That is why PodSwap's TikTok side is built around pacing and mix rather than volume, why the numbers are published rather than hidden, and why the daily ceiling is 40 engagements rather than a figure that would look better in marketing.

Matching by subject and account size means the people watching are plausible viewers of that video. Their attention teaches the system something true instead of something false, and what they say afterwards is worth reading.

It also changes what can honestly be promised. Matched distribution, early qualitative feedback and measurement are deliverable. Views, reach and followers are not, and no service can promise them without lying about who controls them.

Nothing is ever automated, proxied, credentialed or performed on your behalf. Every action is a person opening the app and watching a video, which is both the ethical position and the only version that produces attention worth having.