If you divide the number of visual editor edits by the number of non-bot edits, you get a figure around 10%.
However, based on a number of conversations, I've started to suspect that this may not accurately represent the use of visual editor because many of those non-bot edits may be made using custom tools like Huggle, AutoWikiBrowser, and one-off scripts. Visual editor isn't an alternative to any of those tools; instead, it's an alternative to either of the wikitext editors. So the most accurate comparison would to wikitext editor edits.
This isn't directly possible since T188433 is not done, but proxies are possible. In the past, @Tbayer analyzed the webrequest tables and found that the 2010 wikitext editor accounted for 25% of all edits (including bot edits).
We can improve on this by looking at the ratio between the interfaces found among save events in the Edit event log.
Here, I'll