OBJECTIVES: Neuronavigation is crucial for locating intracranial lesions in neurosurgery. However, it is unaffordable in numerous resource-limited areas. The emerging mobile augmented reality (AR) provides a low-cost alternative to locate lesions, but its accuracy still require improvement before widespread use. This study aimed to explore a novel smartphone AR solution for lesion localization based on a newly developed application and refined reference markers. METHODS: The smartphone AR solution and standard navigation were performed to locate intracranial lesions in 38 patients. The time required for AR and navigation, the deviation between lesion center points identified by AR and navigation, and the ratio of overlap region (ROR) between the lesion locations determined by both methods, were measured, respectively, to evaluate the AR performance in preoperative planning. RESULTS: The average time required for AR was shorter than that for navigation (256.61 ± 69.75 s vs. 454.16 ± 78.85 s, < 0.05), indicating the favorable efficiency of AR. The average deviation and ROR were 3.55 ± 1.71 mm and 75.03% ± 18.56%, which were within the acceptable range of intracranial lesion surgery. The overall accurate localization rate of AR was 81.57%. Moreover, compared to the first stage of this study, the time required for AR and deviation in the second stage were significantly reduced, and ROR was notably increased ( < 0.05). It revealed that with the accumulation of experience, AR efficiency and accuracy were improved. CONCLUSION: The smartphone AR-based solution provides a practical and reliable alternative to locate small intracranial lesions, especially in settings where neuronavigation is unavailable.