River restoration is widely applied, although its effects are poorly understood, and degraded habitats might be difficult to improve. Moreover, there is a lack of monitoring as well as few systematic comparisons of restoration methods. This study presents results of a 4‐year monitoring on River Inn (southern Germany) investigating restoration by gravel or sand addition or embankment removal. The results were compared with reference sites that represent the pre‐restoration conditions. At the landscape scale, we analysed vegetation types based on aerial photographs, whereas at a smaller scale, we undertook vegetation surveys and evaluated species composition, growth, and life form, as well as the proportion of the target vegetation. After 4 years, the data indicated a “negative resilience” of the vegetation back to the state prior to restoration. The structural analysis revealed an extensive spread of reed at expense of bare soil. Thus, the species composition largely regressed to the pre‐restoration conditions, and neither annuals nor other pioneer species showed a long‐term benefit of river restoration. There were differences among the three restoration treatments after 2 years, but no longer after 4 years. However, the river restoration had three positive outcomes: (a) There was a temporary benefit for pioneer vegetation that most likely replenished the seed bank of the respective species, (b) the valuable reed communities showed resilience, and (c) the measures allowed some practical learning as expected for adaptive restoration.