•The case of gradual degradation with unknown parameters is considered.•Remaining lifetime and a preventive threshold are combined with different options.•The long run maintenance cost is derived for unknown degradation parameters.•A Bayesian framework is used for deterioration model update after inspections.•A dynamic procedure for update of decision variables is proposed and assessed. With the development of monitoring equipment, research on condition-based maintenance (CBM) is rapidly growing. CBM optimization aims to find an optimal CBM policy which minimizes the average cost of the system over a specified duration of time. This paper proposes a dynamic auto-adaptive predictive maintenance policy for single-unit systems whose gradual deterioration is governed by an increasing stochastic process. The parameters of the degradation process are assumed to be unknown and Bayes’ theorem is used to update the prior information. The time interval between two successive inspections is scheduled based on the remaining useful life (RUL) of the system and is updated along with the degradation parameters. A procedure is proposed to dynamically adapt the maintenance decision variables accordingly. Finally, different possible maintenance policies are considered and compared to illustrate their performance.