The 450-million-year-old symbiosis between the majority of land plants and arbuscular mycorrhizal fungi (AMF) is one of the most ancient, abundant, and ecologically important mutualisms on Earth. Yet, the evolutionary stability of mycorrhizal associations is still poorly understood, as it follows none of the constraints thought to stabilize cooperation in other well-known mutualisms. The capacity of both host and symbiont to simultaneously interact with several partners introduces a unique dilemma; detecting and punishing those exploiting the mutualism becomes increasingly difficult if these individuals can continue to access resources from alternative sources. Here, we explore four hypotheses to explain evolutionary cooperation in the arbuscular mycorrhizal symbiosis: (1) pseudo-vertical transmission and spatial structuring of plant and fungal populations leading to local adaptation of partners; (2) luxury resource exchange in which plants trade surplus carbon for excess fungal nutrients; (3) partner choice allowing partners to associate with better cooperators; and (4) host and symbiont sanctions which actively reward good partners and punish less cooperative ones. We propose that mycorrhizal cooperation is promoted by an exchange of surplus resources between partners and enforced through sanctions by one or both partners. These mechanisms may allow plant and fungal genotypes to discriminate against individuals employing exploitative strategies, promoting patterns of partner choice. Together these selection pressures provide a framework for understanding the stabilization of mycorrhizal cooperation over evolutionary time.