Very Few Can Distinguish AI Voices from the Real Ones

Article Featured Image

Only one in 50 people can correctly identify artificial intelligence-generated voices compared to the real human ones, a study by cloud communications company Ringover found.

According to the data, 78.3 percent of people, before being tested, felt confident that they would be able to tell the difference between AI and human voices. In addition, almost two-fifths (39.6 percent) believed that their friends and family would be able to tell if they were contacted by a voice mimicking their own. And with banks increasing their use of voice recognition as a security measure, more than two thirds (68.3 percent) of respondents consider their voices to be secure.

In actuality, just one in 50 (2 percent) of people were able to correctly identify all of the celebrity AI voice samples given.

While only 2 percent identified the fake from the real celebrity voice in all five instances, more than a third (35.3 percent) were able to tell the difference between at least one celebrity voice and the AI clone. Oprah Winfrey's was the most easily spotted, with 42.7 percent guessing her real voice correctly, while Prince Harry's proved to be the trickiest, with just 28.5 percent correctly identifying his voice. Others in the survey included Barack Obama (40.4 percent), singer Miley Cyrus (35.5 percent), and actor/comedian Kevin Hart (29.6 percent).

The research also found that those aged 45-54 were the best at detecting AI voices, correctly identifying three out of five voices 22.9 percent of the time, while millennials (25-34 year olds) were the worst, only guessing correctly 12.75 percent of the time. Those aged 65 and older were unable to correctly identify at least three correct voices. Among other cohorts, those ages 35 to 44 were at 17.6 percent, those 18 to 24 were at 13.6 percent, and those 55 to 64 were at 13.5 percent.

After the experiment, almost half (47.6 percent) of respondents felt more concerned about the use of AI, and more than four in five (84.2 percent) were more concerned with how easily people could be fooled into scams by voice generated AI.

"This study highlights how easy it can be for us to be tricked by AI. Voice recognition software could pose a huge potential risk to businesses and customers. Being able to bypass security on phone banking by mimicking other people's voices in order to obtain personal details and access accounts is really concerning," said Renaud Charvet, co-founder and CEO of Ringover. "While this study was conducted listening to celebrity voices, it shows how accurate these pieces of software are at mimicking the voices of others and how misleading they can be."

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues