The debate on whether artificial reefs produce new fish or simply attract existing fish biomass continues due to the difficulty in distinguishing these processes, and there remains considerable doubt as to whether artificial reefs are a harmful form of habitat modification. The harm typically associated with attraction is that fish will be easier to harvest due to the existing biomass aggregating at a newly deployed reef. This outcome of fish attraction has not progressed past an anecdotal form, however, and is always perceived as a harmful process. We present a numerical model that simulates the effect that a redistributed fish biomass, due to an artificial reef, has on fishing catch per unit effort (CPUE). This model can be used to identify the scenarios (in terms of reef, fish, and harvest characteristics) that pose the most risk of exploitation due to fish attraction. The properties of this model were compared to the long‐standing predictions by Bohnsack (1989) on the factors that increase the risk or the harm of attraction. Simulations revealed that attraction is not always harmful because it does not always increase maximum fish density. Rather, attraction sometimes disperses existing fish biomass making them harder to catch. Some attraction can be ideal, with CPUE lowest when attraction leads to an equal distribution of biomass between natural and artificial reefs. Simulations also showed that the outcomes from attraction depend on the characteristics of the target fish species, such that transient or pelagic species are often at more risk of harmful attraction than resident species. Our findings generally agree with Bohnsack's predictions, although we recommend distinguishing “mobility” and “fidelity” when identifying species most at risk from attraction, as these traits had great influence on patterns of harvest of attracted fish biomass.