From the little research I've read about it, the statistical properties of the Voynich Manuscript seem to be much closer to human language than to a text produced by any method that an ancient hoaxer would likely have come up with, but also not quite similar enough to human language that we shouldn't be suspicious of it.
Of course, for any statistical pattern that we might associate with human language, it is possible to design a random process which produces output with that pattern, but the more complex that algorithm is, the less likely it is that someone living 500 years ago would have come up with it, especially when a much simpler process would have been enough to trick their contemporaries.
One interesting example of statistical oddity is the presence of "words" (or "vords") repeated three or four times. That doesn't seem like a very natural sequence to find in a genuine text (especially in an era when documents were slow and expensive to write out longhand), but nor is it something that such an elaborate hoaxer would choose to include either, if they were trying to convince people that the text had meaning.
A perhaps related observation is the lack of crossings out or "mistakes" in the text. This might suggest that the person (or persons) writing the text didn't know or care what they were writing, or alternatively that they held the work in such high regard that they started with a fresh sheet whenever they made a mistake.
Interesting, thank you. I hadn't looked into Rainer Hannig's proposed solution before, but it seems there is a lot of scepticism out there from experts who doubt the "transliterated Hebrew" hypothesis.
I’d love to be wrong though.