Savanna is a type of vegetation influenced by dry and hot climate. Savanna vegetation in Yunnan is distributed in arid and semi-arid valleys of the Hengduan Mountains area, which has an extraordinarily fragile eco-environment. Many artificial forests were planted in these areas in order to restore the vegetation. However, there are few studies on biodiversity conservation and ecological restoration. Ant communities are considered useful indicators of the changing environment and have been used successfully in the evaluation of ecological environments around the world. In order to reveal the bioindicator role of ground-dwelling ants in the savanna of Yunnan, exploring the importance of artificial forests in biodiversity conservation. We investigated the ground-dwelling ants by pitfall traps in plantations and natural vegetation within savanna area, Yunnan, China in 2005 and 2011. The main results were as follows: (1) Ant community composition: 40467 individual ants were collected, representing 5 sub-families, 19 genera, and 47 species. Iridomyrmes anceps was the most common species in well-protected natural vegetation, while Monomorium mayri was the most common one in disturbed natural vegetation and most plantations.(2) Ant abundance and α-diversity: there were no significant differences between the shrub grassland and secondary forest of natural vegetation. Among all vegetation, Azadirachta indica had the greatest abundance of ants and α-diversity, which was close to that of natural vegetation, followed by Eucalyptus spp., and Leucaena leucocephala ranked the lowest. Among all afforestation models, the Azadirachta indica-Acacia mangium plantation had the most α-diversity. Ant abundance was greatest in the Azadirachta indica plantation, and the Eucalyptus spp. plantation was the lowest on both indexes. (3) Community similarity and β-diversity: Only the ant community composition in the Azadirachta indica plantation was similar to that in widely distributed scrub grassland. However, β-diversity in the Leucaena leucocephala plantation was the highest; its βCs value ranged from 0.481 to 0.935; the Azadirachta indica plantation had the lowest diversity, with βCs values ranging from 0.200 to 0.478. (4) Relationship analysis: Both ant species richness and its the ACE value had a significant positive relationship with herb species richness and its ACE value. In total, the most abundant ant species in the ant community in the Yunnan savanna were from Myrmicinae and Camponotus compared with other families and genera, which was similar to that of other savannas in the world. Azadirachta indica and Eucalyptus spp. plantations had a high level of α-diversity that was close to that of the natural forest, which has a positive role in local biodiversity protection. Even with a relative lower α-diversity, the Leucaena leucocephala plantation was a good inhibiting desertification model in an area of extreme degradation with higher β-diversity. All of the artificial forests played an important role in restoring ecological environment in savanna areas. We conclude that the α-diversity in ant communities could be a good indicator of biodiversity, which can reflect the biodiversity situation during vegetation restoration in savanna areas. In order to improve the ecological consequences of artificial forests, more research is needed on the effects of restoration models with mixed tree species and plantation modification.