In multicellular organisms, there is a potential risk that cheating mutants gain access to the germline. Development from a single-celled zygote resets relatedness among cells to its maximum value each generation, which should accomplish segregation of cheating mutants from non-cheaters and thereby protect multicellular cooperation. Here we provide the crucial direct comparison between high- and low-relatedness conditions to test this hypothesis. We allow two variants of the fungus Neurospora crassa to evolve, one with and one without the ability to form chimeras with other individuals, thus generating two relatedness levels. While multicellular cooperation remains high in the high-relatedness lines, it significantly decreases in all replicate low-relatedness lines, resulting in an average threefold decrease in spore yield. This reduction is caused by cheating mutants with reduced investment in somatic functions, but increased competitive success when fusing with non-cheaters. Our experiments demonstrate that high genetic relatedness is crucial to sustain multicellular cooperation.