Most XML is not human-readable. There's just too much line noise, and most of what's emitted has a weird schema that's difficult to parse using Human Brain 1.0.
I don't think it's much worse than CSV, with it's inflexible structure, not allowing for any formatting to be inserted. As for weird schemas, that really is an implementation specific issue. This example was sketched up in minutes and already reads better than most spreadsheets I've seen over the years!
While weird schemas are an implementation-specific issue, the fact remains that most XML that I've seen in commercial settings is exactly as unreadable as I described. It's harder to go against the grain of the status quo when very few people accompany you on the journey.