Corporate Reputation Management in the Era of Artificial Intelligence and Fake News: A Case Study of Safaricom PLC

Abstract

Artificial intelligence (AI) has revolutionised the corporate misinformation space, allowing for the quick generation of synthetic fraud content, deepfakes, and faux announcements that can overwhelm organisations' detection and response capabilities. While both Situational Crisis Communication Theory (SCCT) and Image Repair Theory (IRT) provide the most popular models to guide the handling of reputational crises, both theories were developed before the advent of generative AI and have not been tested systematically against threats posed by AI, particularly in the mobile-money sector, where Safaricom PLC's mobile-money platform M-Pesa dominates the market in Kenya. This study explored typologies of AI-generated misinformation targeting Safaricom, the crisis communication counter-strategies it used to respond, their effectiveness in protecting its corporate reputation and implications for stakeholder trust, and the continued relevance of SCCT and IRT in this context. The study adopted a qualitative, interpretivist approach using a single instrumental case study design, and triangulated data from key informant interviews with 24 respondents (including Safaricom employees, external stakeholders, and experts), documentary analysis of 41 documents, and content analysis of 3,412 social media posts from 2021 to 2026, analysed thematically following the six-step approach by Braun and Clarke. The top four types of misinformation included fabricated mobile-money fraud alerts, fabricated announcements, impersonation accounts and deepfakes. Safaricom responded mostly reactively and denialist in 55% of severe incidents, in line with its internal response-time target. Social media sentiment improved significantly after the official response, but stakeholders perceived the response as not as rapid, especially regarding the longevity of results achieved through closed communication channels. The current practice is limited by the detection lag, misinformation formation in the closed channel and overemphasis on the measures of the process of assessment. The results of the study reveal that SCCT and IRT need theoretical extension to fully capture the dynamics of fast-paced misinformation channels, their speed, and scale in AI-exposed markets, and provide practical tools and strategies to improve corporate crisis communication in these markets.