Antimicrobial resistance is a One Health issue that must be tackled worldwide. In order to implement effective communication strategies in Madagascar, a better understanding must be gained of practices and perceptions related to antimicrobial use at the smallholder farm level. Our study used a semi-qualitative approach, called Q methodology, to identify patterns of opinion on antimicrobial use, or its alternatives, among pig and poultry smallholders and drug vendors in the commune of Imerintsiatosika, in Madagascar. Twenty-nine breeders and 23 drug vendors were asked to rank, respectively, 38 and 45 statements, produced from semi-structured interviews and secondary data, through a 7 grade scale from −3 (totally disagree) to +3 (totally agree) about antimicrobial use, related risks and alternatives. The interview ended with a discussion around extreme statements. The Q-sortings were analyzed by factor analysis and Principal Component Analysis. Regarding antimicrobial use, antimicrobial resistance and alternatives, the breeders and drug vendors were divided according to three discourses: “A: confidence in antibiotics” (respectively, 13 and 6 individuals), “B: belief in alternatives” (7 and 7 individuals), and “C: moderate approach to antibiotic use” (6 and 6 individuals), explaining, respectively, 57 and 60% of total variance. Group A was associated with the use of antibiotics as a preventive measure, poor knowledge of resistance and low trust in alternatives. Group B considered the preventive use of antibiotics to be a major problem for antimicrobial resistance and believed that alternatives, such as vaccines, were useful preventive methods. Group C seemed to have a hazy opinion. The presence of three main points of view offers the possibility to adapt awareness messages. Group B might also be used as a showcase to reduce the amounts of antibiotics used by the two other groups. This study revealed different practices and risk perceptions related to antimicrobial use that must be better characterized and accurately quantified.