Trusting the AI-Voice of the State? Synthetic Orality, Democratic Listening, and Public Authority

This paper examines how AI-generated text-to-speech (TTS) voices shape trust in the interaction with public authorities, focusing on Denmark’s common digital citizen portal Borger.dk in light of the EU Web Accessibility Act coming into force in June 2025 (European Commission). While TTS is implemented to improve access for citizens with difficulties reading on-screen text, it also represents a new institutional voice – a voice that not merely deliver information but also performs the authority of the state.<br/>The paper explores how different synthetic AI-generated voices used for TTS-solutions affect citizens’ perception of authority. Are the voices experienced as trustworthy, caring, competent, and inclusive, or as mechanical, alienating, and hostile? What signs of gender, physical body, age, and emotional state do they express (McDowell, &amp; Ignacio 2023)?<br/>While most concerns about large language models (LLMs) center on textual accuracy and misinformation, the rise of audio-AI where machines are listening and speaking, fundamentally changes the dynamics of trust and mistrust in AI generated content and communication. As I argue in the book Writing with voices, reading with ears: The aural and oral turn in digital communication and media culture (Have, forthcoming) people’s interaction with technology is turning from text input and output, over touch screens, to voice input and output (see also West, Kraut &amp; Chew, 2019 and Rodero, 2018). However, research in how this turn influences digital reading and communication is still sparse.<br/>In Denmark, where digital self-service is a default mode of governance and public trust in institutions is exceptionally high, the move toward synthetic voices in citizen services represents both a technical shift and a cultural risk. Increasing reliance on TTS-tools raises urgent questions of control and access to information as American Big Tech companies are designing “the voices of the state”. Further, text-to-voice technologies accentuate if the text on Borger.dk – and the internet in general – is even suitable for being read aloud? This is generally not the case.<br/>Grounded in the project Resonating Synthetic Voices and Listening Citizens (AIAS-SHAPE, 2024-25), the paper addresses how AI voice systems reflect and reproduce normative biases. Accents, dialects, and speech patterns that fall outside commercial “standard” models may be misrecognized or omitted, contributing to a subtle politics of exclusion. This not only affects accessibility but also legitimacy: if the voice of the state does not reflect the diversity of its citizens, whose authority does it perform?<br/>Drawing on an interdisciplinary theoretical framework based on sound and media theory – including Kate Lacey’s notion of listening publics (2013, 2014), Hartmut Rosa’s theory of resonant democracy (2021), and Walther Ong’s concept of orality (1982) – the paper introduces the concept synthetic orality. It argues that AI generated voices and TTS technology form a critical terrain where institutional trust is negotiated. Hence, when discussing the use of AI tools in the communication between citizens and governments in democratic societies, listening and digital audio-reading should be discussed alongside silent, visual text-reading. <br/>