EVIDENCE 007
YouTube’s own AI assistant says the algorithm is targeting the wrong audience
Sometimes reality forces even a robot to forgot whose side it was supposed to be on...
YouTube Studio’s AI assistant concluded that the recommendation system was serving my videos to the wrong audience and then misreading the resulting lack of clicks as poor content performance.
It says YouTube is placing my videos next to content that does not match my audience — then interpreting the resulting lack of clicks as evidence that the videos are “bad.”
Transcript of the correspondence:
2. Traffic sources: where the algorithm is making a mistake
The main problem is Suggested Videos.
• This source accounts for 70% of all impressions on the channel (about 37K per month).
• But CTR from this source is only 3.4%.
This means YouTube is placing your videos in recommendations next to videos that do not match your audience. The system is trying to “guess” who might be interested in you, but it is doing so poorly.
Because people in those recommendations do not click, the algorithm reaches the mistaken conclusion that the video is “bad” and stops giving it new impressions.
3. The geographic trap
Your main traffic currently comes from Ukraine (15%), Russia (11%) and Uzbekistan (7%). This is the result of Shorts.
If your target audience for long-form videos is the global market (Western Europe, the USA, where Eurodance is active), then the current audience base the system has “learned” from Shorts is simply throwing the algorithm off track.
It looks for viewers for your long-form videos in the same places where it found viewers for Shorts — but it does not find them there.