Adaptation of action potential (AP) duration (APD) to heart's pacing rate is a fundamental electrophysiological feature in cardiomyocytes. Computational models have aimed to recapitulate this phenomenon, in addition to the static restitution, such as in S1–S2 protocol. Following a step decrease in pacing cycle length (CL), APD change occurs in two phases: an initial fast drop in APD, followed by a slow drop phase of up to several minutes until the final steady-state is achieved. Up to date, the modelling community still lack a standard for simulating APD adaptation. In addition, periodically varying pacing CL is typically not included in the in silico protocols, even though their informativeness has been clearly demonstrated. Here, we propose a purposefully simplistic dynamic APD restitution protocol that 1) aims for model independence, 2) links the periodic variability in pacing to physiological heart rate variability, and 3) requires minimal user input in terms specifications. Thus, it can be implemented in a variety of simulation frameworks and potentially enabling cross-context comparisons. We demonstrate the applicability and informativeness of our simplistic protocol by comparing it to the two archetypical approaches: a steady-state (limit-cycle) approach, in which models are paced for on the order of 1000 action potentials to ensure convergence of all state variables, and a quasi–steady‑state approach, in which only tens of action potentials are simulated to approximate rate adaptation. Finally, we show that the periodicity feature of the protocol distinguishes cellular short-term memory properties 1) across a selection of human atrial and ventricular cell models, and 2) under atrial fibrillation-related changes in the phenotype.