To compare the efficacy of therapeutic ultrasound in pain relief and functional recovery in knee osteoarthritis. A comprehensive search of five databases including EMBASE, PubMed, CBM, the Cochrane Library, and Google scholar was conducted to identify relevant studies published between January 1, 2005 and December 31, 2020. Eligible randomized trials were screened for inclusion in this study. Data about the mean change of visual analogue scale (VAS), Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC), and range of motion (ROM) were collected. Fourteen randomized trials covering 1080 patients with treatment durations of 2 to 24 weeks were included. Both pulsed (SMD [CI] = 1.11 [0.86, 1.36], P for heterogeneity < .00001, I2 = 18%) and continuous ultrasound (SMD [CI] = 1.18 [0.78, 1.57], P for heterogeneity < .00001, I2 = 72%) therapy had obvious pain relief effects, and high‐intensity (>1.5 W/cm2) ultrasound seemed more effective (SMD [CI] = 1.34 [0.94, 1.73], P for heterogeneity < .00001, I2 = 35%). In addition, therapeutic ultrasound was also effective in increasing joint function by WOMAC (SMD [CI] = 8.18 [5.88, 10.48], P for heterogeneity < .00001, I2 = 59%). There was a certain degree of heterogeneity due to the differences between the subjects in the study and the ultrasound parameter settings. Our analysis confirmed that both pulsed and continuous ultrasound are effective and safe for pain relief and functional recovery of knee osteoarthritis, especially in high intensity (> 1.5 W/cm2). However, more high‐quality randomized controlled trials will be necessary.