At 6 length, you're already down to the kilometer precision at the equator. At 9 length you're down to 4 meters.
And that's with a dumb grid division, which is very uneven on the globe, the cube projection is probably a better way to deal with it.
Here's the alphabet they chose:
// The character set used to encode the values.
var CODE_ALPHABET_ = '23456789CFGHJMPQRVWX';
There's some information on how they generated it here in the spec[1].
The characters that are used in Open Location Codes were chosen by computing all possible 20 character combinations from 0-9A-Z and scoring them on how well they spell 10,000 words from over 30 languages. This was to avoid, as far as possible, Open Location Codes being generated that included recognisable words. The selected 20 character set is made up of "23456789CFGHJMPQRVWX".
[1]: https://github.com/google/open-location-code/blob/master/doc...
One of the design decisions was to avoid ambiguous characters, so that cuts out at least 9 (I1L 0OB8).
It is almost impossible to make words out of the allowed characters