To evaluate effects of general osteoarthritis (OA) information in addition to a treatment option grid and general practitioner (GP) recommendation to exercise on treatment beliefs and intentions. An online randomized trial of 735 people 45 years old or older without OA who were recruited from a consumer survey network. Participants read a hypothetical scenario about visiting their GP for knee problems and were randomized to the following: i) 'general information', ii) 'option grid' (general information plus option grid), or iii) 'option grid plus recommendation' (general information plus option grid plus GP exercise recommendation). The primary outcome was an agreement that exercise is the best management option (0-10 numeric rating scale; higher scores indicating higher agreement that exercise is best). The secondary outcomes were beliefs about other management options and management intentions. Linear regression models estimated the mean (95% confidence interval [CI]) between-group difference in postintervention scores, adjusted for baseline. Option grid plus recommendation led to higher agreement that exercise is the best management by a mean of 0.4 units (95% CI: 0.1-0.6) compared with general information. There were no other between-group differences for the primary outcome. Option grid led to higher agreement that surgery was best, and x-rays were necessary, compared with general information (mean between-group differences: 0.7 [CI: 0.2-1.1] and 0.5 [CI: 0.1-1.0], respectively) and option grid plus recommendation (0.5 [CI: 0.1-0.9] and 0.9 [CI: 0.4-1.3]). Addition of an option grid and GP exercise recommendation to general OA information led to more favorable views that exercise was best for the hypothetical knee problem. However, differences were small and of unclear clinical importance.