AI video translation is not yet a complete replacement for human translators, according to new research from the University of East Anglia. The study suggests that while artificial intelligence can be highly effective when speed and basic clarity are priorities, human translators remain essential for conveying tone, cultural nuance and natural-sounding speech.
The research was led by Jiseon Han, a lecturer in digital marketing at UEA’s Norwich Business School, who noted that brands expanding rapidly across borders are increasingly experimenting with generative AI. As companies seek to engage global audiences on platforms such as TikTok, Instagram and YouTube, many are asking whether AI can realistically take over video translation tasks traditionally handled by people. “As brands race to reach global consumers, a new question has emerged,” Han said. “Can generative AI truly replace humans in video translation? We decided to put it to the test.”
To explore this, researchers examined how consumers in different countries responded to marketing videos translated by a generative AI tool compared with those translated and delivered by human speakers. The AI system used was HeyGen, which automatically translates language while also modifying voice and lip movements to match the target language. This mirrors the technology many marketers are already trialling in real campaigns.
The study involved two experiments. One focused on Indonesian consumers, while the other involved audiences in the United States and the United Kingdom. Participants watched marketing videos presented by either native human speakers or AI-generated versions. This allowed the researchers to directly compare perceptions of naturalness, comprehension and engagement across different cultural settings.
The results revealed a mixed picture. Viewers consistently rated AI-translated videos as less natural and less convincingly native-sounding than those performed by humans. However, the AI performed better on language comprehension when translating into English, likely because English dominates the data used to train many AI models. Notably, these differences did not affect engagement: participants were just as likely to like, share or comment on AI-translated videos as on human-translated ones.
Overall, the findings suggest that AI video translation already offers clear practical value but has significant limitations. For marketers, AI can be an efficient solution when speed and straightforward messaging matter most. Yet when communication depends on tone, personality and cultural context, human translators remain difficult to replace. As co-author Risqo Wahid from the University of Jyväskylä observed, consumers still notice when something feels slightly off. The study provides a timely snapshot of where AI video translation stands today, highlighting both its growing potential and the enduring importance of the human touch.
More information: Risqo Wahid et al, Generative AI for Video Translation: Consumer Evaluation in International Markets, Journal of International Marketing. DOI: 10.1177/1069031X251404843
Journal information: Journal of International Marketing Provided by University of East Anglia