As a basic rule, you can't use anecdotes to draw assumptions about wide group trends. That my particular Korean friends like the cheesesteak place up the street says nothing about Koreans, cheesesteaks or that restaurant.
This is because at a basic level my friend group just doesn't look like any kind of general population. If nothing else it's geographically restricted, a tiny subset of classes and interests - and likely indirectly education levels and income.
> because interests and ethnic backgrounds are not correlated with each other
They totally are, speaking for populations! This is how de-anonymization typically works. They also correlate with class and income.
And yes, if you target anime, you'll select the Japanese and if you target Webtoons you'll select mostly Korean and Chinese.
Your flaw here is not your language but the direction of your reasoning. You can't reason from the specific anecdata up to population level - "my Black friends like tiktok therefore tiktok must target Black people". That's a "the parking lot is wet, it must be raining" sort of statement.
You CAN reason down. "Tiktok really does a good job showing content from Black influencers, they probably have a large Black audience".