This study introduces a novel method to identify each individual delay in multi-input multi-output linear systems with unknown external delays, and then to apply an adaptive control strategy. The solution is based on a multi-model scheme consisting of a set of tentative delay models running in parallel along with a switching mechanism aimed at progressively updating them. Moreover, the time-varying tentative delays are proved to converge to the actual delays under certain persistent excitation-type conditions on the input signals. The presented approach can be regarded as an online implementation of an heuristic optimisation method known as pattern search, very uncommon in control systems applications. Finally, numerical examples showing the effectiveness of the proposed method are included, as well as the convergence proofs and stability properties of the closed loop. © 2011 The Institution of Engineering and Technology.