Objective To develop evidence-based recommendations for the diagnosis of knee osteoarthritis (OA). Methods The multidisciplinary guideline development group, representing 12 European countries, generated 10 key propositions regarding diagnosis using a Delphi consensus approach. For each recommendation, research evidence was searched systematically. Whenever possible, the sensitivity, specificity and likelihood ratio were calculated for individual diagnostic indicators and a diagnostic ladder was developed using Bayes' method. Secondary analyses were undertaken to test directly the recommendations using multiple predictive models in two populations from the UK and the Netherlands. Strength of recommendation was assessed by the EULAR visual analogue scale. Results Recommendations covered the definition of knee OA and its risk factors, subsets, typical symptoms and signs, the use of imaging and laboratory tests and differential diagnosis. Three symptoms (persistent knee pain, limited morning stiffness and reduced function) and three signs (crepitus, restricted movement and bony enlargement) appeared to be the most useful. Assuming a 12.5% background prevalence of knee OA in adults aged >= 45 years, the estimated probability of having radiographic knee OA increased with increasing number of positive features, to 99% when all six symptoms and signs were present. The performance of the recommendations in the study populations varied according to the definition of knee OA, background risk and number of tests applied. Conclusion 10 key recommendations for diagnosis of knee OA were developed using both research evidence and expert consensus. Although there is no agreed reference standard, thorough clinical assessment alone can provide a confident rule-in diagnosis.