Abstract
In South Africa, citizens are increasingly using social media to monitor and report failures in municipal service delivery. While Citizen-Based Monitoring (CBM) is gaining traction in public sector reforms, its application through social media remains informal, fragmented, and largely unrecognised by municipal authorities. A notable feature of these practices is the use of African indigenous languages, allowing citizens to articulate grievances authentically within their cultural contexts. Guided by Digital Public Sphere theories and the Uses and Gratifications Theory, this study examines how citizens employ digital platforms-particularly through vernacular expression-to participate in public deliberation, seek accountability and fulfil information, expressive, and community-oriented needs. Using a qualitative design, the research combines content analysis of vernacular posts from municipalities with strong social media presences and in-depth interviews with government officials, community-based organisations (CBOs), and community radio representatives. The findings show that indigenous language use amplifies marginalised voices, enhances accessibility, and fosters communal legitimacy around service delivery grievances. However, municipalities often lack the linguistic capacity, digital tools, and institutional frameworks to monitor or respond effectively to vernacular complaints. Meanwhile, community radio and CBOs play intermediary roles without formal support. The study contributes to digital governance, language justice, and participatory democracy scholarship by proposing a framework to formalise vernacular CBM. Key recommendations include developing multilingual digital engagement strategies, formally partnering with community media, and investing in African-language natural language processing (NLP) tools. Recognising vernacular citizen monitoring practices offers a pathway to more inclusive, responsive, and linguistically representative local governance aligned with South Africa's democratic ideals.