The directed-dispersal (DrD) hypothesis, one of the main explanations for the adaptive value of seed dispersal, asserts that enhanced (nonrandom) arrival to favorable establishment sites is advantageous for plant fitness. However, as anticipated by the ideal free distribution theory, enhanced seed deposition may impair site suitability by increasing density-dependent mortality, thus negating the advantage postulated by the DrD hypothesis. Although the role of density effects is thoroughly discussed in the seed-dispersal literature, this DrD paradox remains largely overlooked. The paradox, however, may be particularly pronounced in animal-mediated dispersal systems, in which DrD is relatively common, because animals tend to generate local seed aggregations due to their nonrandom movements. To investigate possible solutions to the DrD paradox, we first introduce a simple analytical model that calculates the optimal DrD level at which seed arrival to favorable establishment sites yields maximal fitness gain in comparison to a null model of random arrival. This model predicts intermediate optimal DrD levels that correspond to various attributes of the plants, the dispersers, and the habitat. We then use a simulation model to explore the temporal dynamics of the invasion process of the DrD strategy in a randomly dispersed population, and the resistance of a DrD population against invasion of other dispersal strategies. This model demonstrates that some properties of the invasion process (e.g., mutant persistence ratio in the population and generations until initial establishment) are facilitated by high DrD levels, and not by intermediate levels as expected from the analytical model. These results highlight the need to revise the DrD hypothesis to include the countering effects of density-dependent mortality inherently imposed by enhanced arrival of seeds to specific sites. We illustrate how the revised hypothesis can elucidate previous results from empirical studies reporting little or no support for the DrD hypothesis, and we suggest its incorporation in designing empirical studies of plant recruitment and in management practices.