Home / Technology / AI Voices Fail Trust Test: Users Reject Synthetic Speech
AI Voices Fail Trust Test: Users Reject Synthetic Speech
11 Mar
Summary
- AI voices are rejected once recognized, harming brand trust.
- Smaller AI firms outperform tech giants in voice quality.
- UK listeners recognize AI voices more than Americans.
A new global study indicates that public trust in AI voices plummets once users recognize them as synthetic, creating a critical issue for businesses employing AI in customer-facing systems. The research, conducted by Vocal Image, analyzed numerous text-to-speech models from major technology firms and specialized developers.
Over 10,000 participants evaluated various AI voices without prior knowledge of their artificial origin. Key findings highlight a strong negative correlation between detecting an AI voice and user acceptance. Differences in quality were stark, with top-performing models rated significantly higher than lower-ranked ones.
Notably, Chinese startup MiniMax was ranked highest by both UK and US listeners, while major tech companies like Google, Amazon, and Microsoft lagged. The study also observed that UK listeners were more likely to identify AI voices than Americans, though European listeners showed greater overall acceptance.
Nick Lahoika, CEO of Vocal Image, stated that choosing the wrong AI voice provider is becoming a brand liability, emphasizing that "people still don't trust bad AI voices." He suggested that Big Tech might increasingly rely on acquisitions to bridge the gap in specialized voice quality, as large companies struggle to achieve the necessary precision across diverse use cases.
