OBJECTIVES: Exercise is an effective and commonly prescribed intervention for lower limb osteoarthritis (OA). Many unanswered questions remain, however, concerning the practical delivery of exercise therapy. We have produced evidence-based recommendations to guide health-care practitioners. METHODS: A multidisciplinary guideline development group was formed from representatives of professional bodies to which OA is of relevance and other interested parties. Each participant contributed up to 10 propositions describing key clinical points regarding exercise therapy for OA of the hip or knee. Ten final recommendations were agreed by the Delphi technique. The research evidence for each was determined. A literature search was undertaken in the Medline, PubMed, EMBASE, PEDro, CINAHL and Cochrane databases. The methodological quality of each retrieved publication was assessed. Outcome data were abstracted and effect sizes calculated. The evidence for each recommendation was assessed and expert consensus highlighted by the allocation of two categories: (1) strength of evidence and (2) strength of recommendation. RESULTS: The first round of the Delphi process produced 123 propositions. This was reduced to 10 after four rounds. These related to aerobic and strengthening exercise, group versus home exercise, adherence, contraindications and predictors of response. The literature search identified 910 articles; 57 intervention trials relating to knee OA, 9 to hip OA and 73 to adherence. The evidence to support each proposition is presented. CONCLUSION: These are the first recommendations for exercise in hip and knee OA to clearly differentiate research evidence and expert opinion. Gaps in the literature are identified and issues requiring further study highlighted.