Maize, the second most globally important staple crop after wheat, originated in Mexico, where it is typically grown as part of a set of associated crops and practices called the milpa system. This ancient mode of production is practiced today in ways that vary by cultural context and agro-environment. Milpas generate private economic value, in terms of food security, diet quality and livelihoods, for the two-million farm households who manage them. Furthermore, milpas generate public economic value by conserving agrobiodiversity, especially that of maize landraces, which have the potential to contribute unique traits needed by plant breeders for future crop improvement. In this way, milpas contribute to global food security in maize. However, the sustainability of the milpa system has been threatened by off-farm employment opportunities, long-distance migration, the increasing commercialization and intensification of maize production. Most recently, the milpa system has been negatively impacted by the contamination of maize landraces by genetically modified (GM) maize, cultivation of which is currently prohibited in Mexico. Here, we employ a choice experiment to estimate Mexican farmers’ valuation of three components of agrobiodiversity (crop species richness, maize variety richness and maize landraces), and examine their interest in cultivating GM maize. Choice experiment data, household level social, economic and demographic data, community level economic development data, and information on milpa production characteristics, and farmers’ attitudes and perceptions with regards to GM food and crops were collected from 420 farm households across 17 communities in three states of Mexico. Using these data, we analyzed the heterogeneity of farmer preferences using a latent class model, which can be used to simultaneously identify sample segments having homogenous preferences for milpa attributes, as well as farmer characteristics affecting preferences. We further identified the characteristics of farmers who are most likely to continue growing maize landraces and managing milpa systems, as well as those least likely to accept GM maize. Specifically, we identified three distinct segments of farmers: (i) Landrace Conservationists derive the highest private economic value from continued management of landraces and the highest economic loss from the possible adoption of GM maize. These farmers are young, dislike GM foods and crops, and are mainly located at the Oaxaca site, where transgenic constructs have been found in maize landraces. (ii) Milpa Diversity Managers derive the highest economic value from managing all of the agrobiodiversity components of the milpa, and suffer fewer losses from management of GM maize. These are older farmers, who are curious and like to experiment with maize varieties. (iii) Marginalized Maize Producers derive little value from crop species and maize variety richness, receive minimal value from maize landraces, and also experience the smallest negative impact from the adoption of GM maize. These farmers are located in the most isolated communities, have the lowest level of productivity, and oversee the largest milpa areas. They are also the most tightly integrated into the maize output markets. These novel findings have implications for debates concerning the adoption of GM maize in Mexico and its associated costs and benefits, as well as for the design of targeted, cost-effective conservation programs on farms