Disorders of communication, social relationships, and psychomotricity are often characterized by cognitive impairment, which hinders daily activities and increases the risk of falls. This study aimed to evaluate the efficacy of an animal-assisted therapy (AAT) program in an institutionalized geriatric population with cognitive impairment. The variables evaluated included level of communication and changes in gait and/or balance. We performed a two-arm, parallel controlled, open-label, nonrandomized cluster clinical trial in two nursing home centers from an urban area. Patients in the two centers received 12 weekly sessions of physiotherapy, but the experimental group included AAT with a therapy dog. The study included a total of 46 patients (23 Control Group [CG], 23 Experimental Group [EG]) with a median age of 85.0 years. Of these, 32.6% had mild–moderate cognitive decline (Global Deterioration Scale of Reisberg [GDS] 2–4) and 67.4% severe cognitive decline (GDS 5–6). After the intervention, patients in the CG and EG showed a statistically significant improvement in all the response variables. When comparing both groups, no statistically significant differences were found in any of the Tinetti scale results (measuring gait and balance). However, the communication of patients in the EG, measured on the Holden scale, showed a statistically significant greater improvement postintervention than that of patients in the CG. AAT can be useful as a complementary, effective treatment for patients with different degrees of cognitive decline.