This study was designed to determine whether feedback from welfare assessments, using the Horse Welfare Assessment Protocol, affected actual horse welfare in 21 stables. After the first assessment, stable managers in the high feedback (HF; n = 10 stables) group were supplied with extensive information and support regarding the welfare measures and relevance of the results. The low feedback (LF; n = 11 stables) group only received the results without additional information. Upon re-assessment, six months later, no significant changes were seen in the stable overall (SO) score in either group. Significant changes occurred in individual measures; in the HF group more fresh-air inlets were open but water drinker function and ocular discharge deteriorated. In the LF group, the feeding troughs were cleaner but mane and tail condition deteriorated. Both groups had cleaner water troughs and less equipment chafing but the sum of relative air humidity (RH) and temperature (T) deteriorated. Significant decreases occurred in the stable welfare issues (SWI) score; the HF group decreased from 93.3 to 72.0 and the LF group from 113.3 to 91.3. There were also non-significant changes; in the HF group, 71 measures and five stables improved while 63 measures and five stables (50%) deteriorated. In the LF group, 65 measures and seven stables improved while 62 measures and four stables deteriorated. The observed improvements in both groups suggest that assessment alone (with no detailed feedback) might raise awareness but we cannot yet conclude whether or not the type of feedback affects overall horse welfare.