genetic algorithms for NP-hard problems

for NP-hard problems, are GAs the same as random bruteforce (or possibly even worse)? or should GAs always be used instead of random bruteforce, just in case there ARE partial solutions, or can the partial solutions make the GA slow down?

to put it simply, GAs or random search when the problem is chaotic?

>>2
yeah, in my implementation the gene is randomized if the same gene is received from both parents to prevent genetic drift, but from my understanding with this approach the GA can still get stuck "orbiting" a hole, so it will follow the punctuated equilibrium (when it finally escapes the orbit, it enters another orbit). are those holes mostly a stumbling block, so random search is better? or do they help the GA to map the possible patterns found?
I have heard of GAs being surprisingly successful for NP-hard and NP-complete problems that couldn't be found an easy solution for.
I assume it's unknowable to determine if GAs will  be better than random search, so I always have to try and see?