Automatic ozone measuring devices have been operating continuously on board the five long-range aircraft of the Measurement of Ozone and Water Vapor by Airbus In-Service Aircraft (MOZAIC) program since September 1994. This paper presents the main characteristics of the ozone system and the procedures followed to ensure its accurate calibration over long durations. Measurement accuracy was estimated at ±[2 ppbv + 2%], but much better in-flight levels were in fact observed: average discrepancy (between different devices) ranging from 1 ppbv at tropospheric concentrations to a few ppbv at stratospheric concentrations. This demonstrates the ability of the MOZAIC ozone data to produce accurate and reliable ozone climatologies. A 2-year ozone climatology (1994–1996) generated from MOZAIC data collected at between O and 12 km altitude was compared to longer and older measurements made at eight stations of the Ozone Sounding Network (OSN): Hohenpeissenberg, Wallops Island, Tateno, Palestine, Pretoria, Goose Bay, Biscarosse, and Poona. Despite the different nature of the programs (techniques, platforms, sampling frequencies, spatial distribution, and operation periods), the OSN and MOZAIC climatologies were found to show a reasonably high level of agreement. Mean concentrations derived from ozone sondes are about 3 to 13% higher than those obtained by the MOZAIC program in the free troposphere, in a similar geographic location. These differences are within the range of uncertainty of the two techniques. Larger discrepancies observed in the boundary layer and in upper layers are explained by the influence of local pollution and the distance between measurements, amongst other factors, limiting the reliability of comparisons. A comparison of OSD and MOZAIC data at Hohenpeissenberg/Frankfurt and Wallops Island/New York, over an overlapping period (1994–1995), shows good agreement in the free troposphere (800–300 hPa), no detectable bias for Hohenpeissenberg/Frankfurt, when taking into consideration the various causes of discrepancies (Dobson normalization, ozone geographical variations). Indeed, the results of this analysis support the hypothesis that it is not advantageous to scale the ozone sonde data to the overhead ozone column; the scaling appears to cause overestimation of the tropospheric O3 concentrations, by about 3–6% at Hohenpeissenberg, and to cause more scatter in the sonde-MOZAIC differences. The correspondence between the OSN and MOZAIC climatologies obtained in very different conditions demonstrates that they are representative of the atmosphere and that, being complementary while each retains its own advantages, they are therefore both useful for validation studies.