This work unveils design flaws within most metaheuristics, with a specific focus on issues associated with the arithmetic optimization algorithm (AOA). Despite being a simple metaheuristic optimizer inspired by mathematical operations, AOA holds promise for addressing complex real-world applications. However, a thorough analysis of its search mechanism reveals a heavy dependence on problem bounds for the quality of solutions obtained by AOA. Additionally, discrepancies between algorithm descriptions and implementations in AOA can mislead users and impede progress within the metaheuristic community. Experimental simulations conducted on various standard benchmarks including their shifted versions indicate a structural bias in AOA, leading to artificially high accuracy in fitting standard test functions but poor performance when applied to shifted benchmarks. Finally, we give a critical cause analysis and conclude this article by highlighting valuable research avenues in this field.