If that documentation serves another human-operated querying AI it already provides value as a cached work that prevents another AI from re-extracting that knowledge from the code (burning time & tokens in the process).

Another question is keeping those docs up to date...