AI-DRIVEN CRISIS COMMUNICATION AND EMERGENCY RESPONSE: OPTIMIZING NONPROFIT DIGITAL OUTREACH DURING NATIONAL DISASTERS

Abstract

Nonprofit organizations occupy a pivotal position in national disaster communication, often serving as the most accessible and trusted intermediaries for at-risk populations. Yet their ability to deliver timely, actionable, locally relevant, and credible information is constrained by limited staffing, multilingual audiences, uneven digital capacity, and rapidly evolving situational demands. Recent advances in artificial intelligence (AI)—including automated triage and routing, conversational agents, machine translation, summarization, and rumor-detection pipelines—offer new affordances for reducing message latency, increasing message personalization, and strengthening credibility cues during high-uncertainty events. This study systematically evaluates whether, and through what mechanisms, AI-enabled communication tools enhance nonprofit digital outreach effectiveness during national disasters. Using a quantitative, cross-sectional, multiple–case design (N = 236 staff respondents across five nonprofit organizations), the study operationalizes AI Adoption Intensity, Message Relevance, Public Trust, Digital Outreach Effectiveness, Digital Readiness, and Disaster Severity with validated five-point Likert scales demonstrating strong reliability (α = .84–.92), composite reliability (CR ≥ .88), and convergent validity (AVE ≥ .59). Descriptive analyses indicate substantial variance in AI adoption (M = 3.22, SD = 0.86) and readiness (M = 3.35, SD = 0.83), providing analytic leverage to examine their relationships with outreach outcomes (M = 3.81, SD = 0.67). Hierarchical regression results show that AI Adoption Intensity is positively associated with Digital Outreach Effectiveness (β = .23, SE = .05, p < .001), improving model fit by ΔR² = .11 after accounting for organizational controls. When Message Relevance and Public Trust are included as theoretically proximal predictors, both emerge as strong determinants of effectiveness (MR: β = .36, p < .001; PT: β = .22, p < .001). Bootstrapped mediation analyses (5,000 resamples) confirm that AI’s association with outreach effectiveness is partially transmitted through Message Relevance (β_indirect = .14; 95% CI [.09, .21]) and Public Trust (β_indirect = .08; 95% CI [.04, .14]), yielding a total indirect effect of .22. The residual direct effect of AI remains significant (β = .09; 95% CI [.01, .17]), indicating partial—but not full—mediation. Moderation models further reveal that Digital Readiness amplifies AI’s marginal benefits (interaction β = .12, p < .01), with the AI → effectiveness slope increasing from non-significant at low readiness to β = .31 (p < .001) at high readiness. Conversely, Disaster Severity attenuates AI’s returns (interaction β = −.10, p < .05), as extreme operational strain and verification bottlenecks reduce the translation of algorithmic speed into perceived clarity and trust.