Hong Kong AI fools 90% of consumers in voice tests

Survey shows most callers mistake synthetic speech for staff creating identity risk.

Nine in 10 Hong Kong consumers cannot identify artificial intelligence (AI)-generated voice clips from human agents, according to a Twilio survey.

The platform defined AI identity crisis as voice bots sound indistinguishable from humans, which should prompt regulatory intervention, the report said.

Twilio’s study found that Hong Kong consumers have low tolerance when their expectation with AI falls short, with only 35% of consumers satisfied with their AI-supported customer service.

“In markets like Hong Kong with dominant financial and insurance sectors, regulatory intervention is therefore a certainty,” the firm said.

“To maintain brand trust, companies will need clear, explicit identification, such as ‘I am Ruby, an AI assistant,’ to avoid the reputational fallout of synthetic deception,” it added.

Local consumers demand trust and clear control, according to Twilio’s State of Customer

Engagement Report 2025, with half (50%) of Hong Kong consumers preferring to speak to a human agent when it is needed.

About 83% of consumers want control over how brands communicate with them, rather than having AI or agents automatically assume their preferences.

Meanwhile, 52% of local users prefer faster service and are willing to sacrifice service quality for speed, the platform's digital patience study showed.

This is a trend that the study predicted will continue in 2026.

To improve AI tools, brands will decisively move towards modular, Bring-Your-Own-LLM (BYO-LLM) architectures that are easier to update, cheaper to run, and better equipped for nuanced local language, Twilio said.