Other ideograph languages (e.g. Chinese) aren’t so bad when you do this, because enough information is captured per ideograph that losing ordering doesn’t actually change meaning that much. But Japanese is especially bad, because it often “reverts” to the hiragana or katakana alphabets for long strings of words, while still not putting any explicit word-break markers between saidwords. So you end up indexing a “bag of letters.” You can imagine the uselessness of such results.
Or, to put that another way: Japanese-language search “works” on English-language sites when you’re searching a term expressed in kanji. But when the term you want turns out to only have a katakana representation, it’s mostly not going to work.
And yes, Google themselves know all this and have solved all these problems for Google Search long ago. Apparently, the relevant tech never made it into YouTube.