Abstract The study of morphology in galaxies offers a convenient and quantitative method to measure the shapes and characteristics of galaxy light distribution that reflect the evolutionary history. For active galactic nucleus (AGN)-host dwarf galaxies, however, there is a lack of detailed studies of their morphologies. In this work, we compile a relatively large sample (∼400 members) of local AGN-host dwarf (M ⋆ ≤ 109.5 M ⊙ and z < 0.055) galaxies selected via various methods. We use the grz bands images from DESI DR10 and the Python package statmorph to measure nonparametric coefficients. We also carry out visual inspection with the assistance of deep learning to classify these galaxies into early-type galaxies (ETGs), late-type galaxies (LTGs), and mergers, and find that about 37%, 44%, and 13% of the total sample sources are ETGs, LTGs, and mergers, respectively. In comparison to normal dwarf galaxies, AGN-host dwarfs have a higher probability to be LTGs, and a lower merger rate, indicating that mergers/interactions are not the primary driver of AGN activities. Among the subsamples selected with different methods, the Baldwin, Phillips & Terlevich sample has the highest fraction of ETGs, the variability sample consists of the largest fraction of LTGs, and the mid-IR sample contains the most mergers.