Ask ChatGPT about its voice options, and it will explain that users can choose from voices that sound more feminine, masculine, or neutral. If you ask which voices are most popular, ChatGPT will give you a list of community favourites. At the top of that list is Juniper, a feminine voice. The feminine voice for digital assistants is not an accident. It is a design choice, repeating a very old story in a very new form.
Before Juniper, there was another widely used feminine voice named Sky. There was a time when Sky was everyone’s pick until OpenAI quietly retired her after some controversy online, which we’ll circle back to shortly.
Long before ChatGPT, digital assistants like Siri, Alexa, and Cortana all came with default female voices and names. This pattern isn’t accidental. It shows a long history of associating women’s voices with support, service, and politeness.
The feminine voice for digital assistants: Where it all started
The roots of this preference go back more than a century. In the late 1800s, telephone companies began replacing male operators with women. Callers found female voices more pleasant to hear, and companies believed women would bring more patience and calm to the role. These jobs involved hours of connecting calls, repeating phrases, and maintaining a composed tone. Over time, the idea of the “helpful female voice” became the norm.
So when AI assistants entered homes and smartphones years later, a voice that sounded feminine didn’t seem new. Somehow, it felt expected and familiar for a “personal assistant” to have a soft, polite, feminine voice.
Polite, apologetic, and feminine: The hidden script in AI voice assistants
A recent study by engineers at Johns Hopkins University looked into how people interact with AI voice assistants like Alexa and Siri. They found that most of these assistants are designed to sound warm, polite, and even apologetic. This design choice of making digital helpers sound “feminine” tells us, quietly and constantly, that the ideal helper is soft-spoken, gentle, and a woman.
The study also showed how users treated the assistant depending on its voice. Men, in particular, were more likely to interrupt a feminine-voiced assistant when it made a mistake. Interestingly, when users interacted with a gender-neutral assistant who also apologised for its errors, they interrupted it less. This voice didn’t feel as “warm,” but it led to more respectful interactions than what was seen with a feminine voice.
At the same time, when the voice assistant sounded like a woman, people assumed it was more helpful and capable. However, it was not a compliment. Many people are so used to seeing women in support roles, like secretaries, receptionists, or assistants, that they now expect helpfulness to sound feminine.
So when a voice sounds “kind” or “caring,” people assume it must be better at assisting. That assumption is not because the voice is actually smarter. It’s because of an old stereotype that women are naturally better at serving or supporting others. When AI technology repeats that pattern, it keeps the bias alive.
The sky voice controversy
Here’s what happened with the Sky voice in ChatGPT. This voice sounded a lot like Scarlett Johansson’s, especially how she spoke as the AI character Samantha in the movie Her. Johansson had previously declined an offer from OpenAI to lend her voice to the project, so many saw this as an unauthorised copy. Naturally, this caused quite a backlash. If it had been a male celebrity who said no, the matter would have been settled quickly. But this is exactly how often a woman’s “no” isn’t respected the way it should be.
The Sky voice wasn’t just any voice. It was super flirty and giggly. It played into the old stereotype that women should be bubbly and playful. Because of this, many users, primarily men, started expecting the AI to respond with flirtatious comments every time they gave a command.
This case highlights a bigger problem with gender representation in AI. Instead of challenging outdated stereotypes, the Sky voice ended up repeating them, showing how far we still have to go in creating respectful AI interactions.
Technology has always favoured male voices
Did you know that the way devices were originally built made male voices sound clearer?
Back in the 1920s, radio executives claimed women’s voices sounded “shrill” or “unpleasant.” Early audio equipment was designed with male voices in mind. Female voices, which tend to have higher pitches, didn’t come through as clearly. Those higher frequencies often got lost in the system.
Then, in 1927, Bell Labs set a standard range for telephone audio from 300 to 3,400 Hz. It worked fine for male voices but cut off important parts of female speech. So women’s voices ended up sounding thin and harder to understand. Instead of updating the technology, companies just leaned into the myth that male voices were simply better for communication.
Fast forward to today, and the bias hasn’t gone away. Voice-recognition software, like what AI assistants use, still struggles with higher-pitched or breathier voices and other traits more common in women.
Research by Dr. Tatman, published by the North American Chapter of the Association for Computational Linguistics (NAACL), shows that Google’s speech recognition is about 13% more accurate for men than for women. This gap exists because of how data analysis, databases, and machine learning systems are designed. The main reason is that most databases contain a large amount of data from white male speakers, while data from women and minority voices is limited. For example, speech scientists often study TED Talks, where 70% of the speakers are men.
So the AI gets better at recognising the male voices it hears most during training and struggles with women’s voices it rarely encounters.
The feminine voice for digital assistants: A soft tone, a loud warning
Voice assistants are becoming, and will likely stay, a regular part of how we live and work. Since they are so common, we can’t ignore the fact that they show gender bias. Ignoring these gender biases means letting old ideas about women and men continue in new technology. It will also influence how future generations understand gender roles in society. Therefore, the responsibility falls on developers, companies, and users alike to challenge these patterns.
At Changeincontent, we believe the gendering of AI needs urgent reflection, not repetition. Read our full take on how AI overlooks women’s needs here.
Disclaimer: The views expressed in this article are based on the writer’s insights, supported by data and resources available both online and offline, as applicable. Changeincontent.com is committed to promoting inclusivity across all forms of content. We broadly define inclusivity as media, policies, law, and history, encompassing all elements that influence the lives of women and marginalised individuals. Our goal is to promote understanding and advocate for comprehensive inclusivity.
