In the paper the random-seeking counterparts of the stochastic approximation procedures are treated. Sufficient conditions for the convergence with probability I of the analogues of Kiefer-Wolfowitz'a and Kabian'a algorithms are presented and discussed. These conditions are obtained by applying a unified and general approach (also presented in the paper) to proving the convergence of minimization algorithms defined by stochastic difference equations. Some advantages of random-seeking methods are described.