The scale and extent of global urbanization are unprecedented and increasing. As urbanization generally encroaches on natural habitats and the urban ecological footprint reaches far beyond the city limits, how urbanization affects biodiversity has received increasing attention from the scientific community. Nonetheless, the comprehensive syntheses of urbanization consequences for biodiversity, including diverse taxonomic groups, across multiple spatial scales and spanning a wide gradient range of urbanization intensity are still insufficient. Here, based on the urban-rural gradient approach, we assessed the effects of urbanization on Chinese mammal and amphibian richness across the entire urbanization gradient (i.e., urbanization level from 0 to 1) at the national, regional and urban agglomeration scales. We used the global mammal and amphibian distribution data along with corresponding background climate, habitat conditions and socioeconomic activities data for analysis. Our results revealed a detailed and diverse pattern of Chinese mammal and amphibian richness along the entire spectrum of urbanization gradient across three spatial scales. And an approximately monotonic decrease only existed in certain urban agglomerations. The imprint of urbanization on mammal and amphibian richness were largely masked by the overall primacy of background climate at the national and regional scales. As the scale of analysis shifting from the country to urban agglomerations, urbanization-associated variables and locally specific limiting factors started to play important roles in driving the richness patterns. Moreover, the environmental Kuznets curve hypothesis can explain the relationship between biodiversity pressure and urbanization activities in certain Chinese urban agglomerations. However, the findings of urbanization effects on biodiversity using the urban-rural gradient analysis should be interpreted with caution because many possible driving forces simultaneously present along the urban-rural gradient and are very challenging to attribute.