Objectives: This study aimed to establish evidence for effective (Augmentative and Alternative Communication) AAC intervention approaches for patients with dementia of Alzheimer’s type (DAT) by examining whether the severity of dementia and the symbol type used in AAC tools affect symbol selection. Additionally, we aimed to investigate the relationship between symbol selection ability and language and cognitive abilities. Methods: Twenty-one normal elderly individuals, 17 patients diagnosed with mild DAT, and 20 patients with moderate to severe DAT were recruited. A symbol selection task was conducted using the Augmentative and Alternative Communication Symbol Assessment Tool (AAC-SAT). Participants selected target symbols on the screen consisting of graphic symbols, photos, and written words in response to auditory prompts given by the examiner. Data from the tasks were statistically analyzed for accuracy and response time, and correlation analyses were conducted for cognitive and language abilities. Results: The moderate to severe DAT group showed significantly lower accuracy and longer response times in symbol selection compared to the mild DAT group and normal elderly group. Upon examining each group separately, both mild and moderate to severe DAT groups showed no symbol type that was advantageous for symbol selection. However, the rapid response to photos observed in mild DAT group disappeared in the moderate to severe DAT group. Additionally, the symbol selection ability of DAT patients significantly deteriorated as language and cognitive abilities declined. Conclusion: Appropriate symbol types for AAC aimed at facilitating effective communication considering the language and cognitive abilities of DAT patients were suggested, and the findings of this study were compared with the claims made by previous research.