In data‐driven controller tuning methods based on a model matching problem, we sometimes suffer from appropriately selecting a reference model. In order to avoid this difficulty, the estimated‐response‐based controller tuning methods are recently studied. This paper proposes a new estimation method of the closed‐loop response and a controller tuning method using the estimated closed‐loop response. The proposed estimation method is applicable even to unstable systems and deals with the transient input/output data unlike conventional response estimation methods. Moreover, the proposed estimation method estimates the closed‐loop response with a non‐minimum phase or an unstable controller by introducing two fictitious signals. For improvement of estimation accuracy, the proposed method considers to reduce influences by noise and a certain type of disturbance. The controller tuning problem is formulated without any reference model and any information on the plant while imposing time‐domain constraints, overshoot, rise time, settling time, and input limitation. The effectiveness of the proposed method is confirmed by numerical and experiment examples. © 2021 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC.