At the start of life, the origin of a primitive genome required individual replicators, or genes, to act like enzymes and cooperatively copy each other. The evolutionary stability of such enzymatic cooperation poses a problem, because it would have been susceptible to parasitic replicators that did not act like enzymes but could still benefit from the enzymatic behaviour of other replicators. Existing hypotheses to solve this problem require restrictive assumptions that may not be justified, such as the evolution of a cell membrane before the evolution of enzymatic cooperation. We show theoretically that, instead, selection itself can lead to replicators grouping themselves together in a way that favours cooperation. We show that the tendency to physically associate with others and cooperative enzymatic activity can coevolve, leading to the evolution of physically linked cooperative replicators. Our results shift the empirical problem from a search for special environmental conditions to questions about what types of phenotypes can be produced by simple replicators.