This article examines the evolution of crisis communication theory from pre-internet models of institutional response to conditions shaped by generative artificial intelligence. The study addresses a growing mismatch between inherited crisis communication principles and a communication environment in which reputational threats can be produced, amplified, personalized, and interpreted through AI systems. The objective is to explain how crisis communication theory has changed across three stages: the pre-internet period, digital adaptation, and the age of generative AI. The methodology relies on comparative source analysis, conceptual synthesis, typologization, and analytical generalization of ten recent scholarly publications in crisis communication, public relations, misinformation studies, and human-machine communication. The results identify a shift from event-centered response logic to dynamic, adversarial, and machine-mediated crisis conditions. The article proposes a revision path based on anticipatory monitoring, human-machine response governance, authenticity infrastructure, and adaptive trust repair. Its practical value lies in a decision model for organizations facing AI-generated misinformation, synthetic evidence, and algorithmically accelerated reputation crises.