Ideally, this would be designed in two parts: separate the file structure from the metadata discovery mechanism.
I personally want a file structure managed by the OS. Let me make folders and nested subfolders to whatever structure I prefer.
Then make the metadata discovery slightly more manual. Click a media file, click a hypothetical "add metadata" button, and then a simple search box with "is this your movie?" and click apply to import metadata from a search result. easy peasy.
The UI is clearly meant to resemble a typical media app but falls short if the end user prefers, for example, foobar2000's UI.