Macrobenthos is an important part in mangrove ecosystem, and plays a connection link in energy flow and matter cycling. Its community structure can reflect the level of mangrove forest rehabilitation by bottom sediment condition, water quality, vegetation and environmental pressure and so on. To reveal the effects of mangrove rehabilitation on macrobenthic communities, a seasonal investigation (from October 2010 to May 2011) on the macrobenthic assemblages was performed in six habitats, including four artificial Kandelia obovata forests with different ages, Spartina alterniflora (Sa) marsh, and naked beach (NB) in Ximen Island in southeastern China. The four mangrove forests were 1-year old replanted mangrove forest (M1), 4-years old replanted mangrove forest (M4), 8-years old replanted mangrove forest (M8) and 50-years old replanted mangrove forest (M50). To avoid the effects of tidal elevation, all the six habitats had the same elevation. A total of 24 macrobenthic species were recorded, including Polychaeta, Mollusca, Crustacean and others. The members of Mollusca and Crustacean were dominant and dominant species were found to be different among various habitats. Phascolosoma esculenta and Littoraria melanostoma were the dominant species in M50, while Assiminea latericea, Batillaria zonalis, and Cerithideopsilla cingulata were the dominant species in the others. Young mangrove forests were dominated by surface adhering group, while the older mangrove forest (M50), NB and Sa were dominated by infauna. Moreover, it revealed that caving animals appeared only in mature ecosystems which are more stable, such as NB, Sa and M50. There were differences in macrobenthic assemblages among the six habitats and M50 showed the highest stability in terms of ecological function as indicated by the highest abundance and species diversity of macrobenthos. This result was different from those of previous researches, which suggested a negative relationship between the abundance and diversity of macrobenthos and the mangrove development. This difference was probably due to the differences in sediment and soil development. Compared with typical natural mangrove forest, M50 had less litter and semi-decomposed litter in the sediments. Furthermore, the sediment in M50 lacked the typical smelly and black-color appearance which is typical of natural mangrove forest. Therefore, our results suggested that the sediment of M50 in Ximen Island developed slower than those of natural ones, which may be due to the limited planting scale and the larger tidal range in Ximen Island. Generally, further researches focusing on the linear relationship between the macrobenthos biodiversity and the ecosystem stability, especially its applicable temporal and special scale are required.