This paper evaluates the acoustic performance of AI-generated exhibition hall designs using room-acoustic simulation. Four initial design alternatives were generated using an AI design tool, with different zoning layouts and partition arrangements. The alternatives were assessed using reverberation time, speech clarity, sound pressure level, and sound strength. The results showed that the first four alternatives had excessive reverberation and weak alternative speech clarity, although Alternative 3 provided the most suitable layout for further improvement. A second AI prompt was then developed by adding sustainability, zero-maintenance, and biophilic design requirements. Based on this prompt, the AI selected Alternative 3 and modified its partitions into acoustic moss wall partitions. The final alternative showed a clear acoustic improvement, especially in reducing reverberation and sound build-up while strengthening the architectural identity of the hall.