A pseudo-plant contains a nominal model of the controlled process and combines suitable signals to form a pseudo-output which equals the true output if the model is exact. Robustness against unstructured uncertainty is improved using a pseudo-plant, and in adaptive control the parameter estimates become less sensitive to the input frequency. By introducing an appropriate gain term in the pseudo-plant design, the norm of the effective unstructured uncertainty can be made arbitrarily small. It is shown that the approach can handle a larger unstructured uncertainty which satisfies the established robustness result in non-adaptive and adaptive pole-placement control. Simulation results are presented.