Masculine or feminine? Tailoring AI voices for enhanced ad performance

You’re watching YouTube, and an ad pops up for a BBQ grill.

A gruff, deep voice tells you the benefits of the GrillMaster 3000. It’s the stereotypical “manly voice” perceived as congruent with the product.

Then, a soft, feminine voice implores you to purchase a spa package in the next ad.

But are you actually more likely to act on an ad for a stereotypically masculine product when the voiceover sounds masculine, and vice versa for feminine product categories?  

Recent research at the University of St. Gallen in Switzerland tested this assumption and found that congruency toward stereotypically masculine or feminine products could significantly improve advertising performance with AI voiceovers.

Journal of Interactive Marketing, Volume 0. “The Power of AI-Generated Voices: How Digital Vocal Tract Length Shapes Product Congruency and Ad Performance”
Background

In AI-generated voices, digital vocal tract length—the distance from vocal folds to lips—can be adjusted, affecting the perceived qualities of the voice, such as masculinity or femininity. 

VTL is mainly associated with voice timbre; Longer VTL typically corresponds to a more masculine voice, while shorter VTL is associated with a more feminine voice. 

The findings are significant for any marketer who runs ads with AI voiceovers and, with less reliability, voiceovers in general. 

About the study

The Journal of Interactive Marketing study explored how these changes in AI voices can impact advertising performance. 

The research was conducted by William H. Hampton, a Senior UX Research Manager at TikTok, Fotis Efthymiou, a doctoral candidate at the University of St. Gallen, Christian Hildebrand, Professor of Marketing Analytics at the University of St. Gallen, and Emanuel de Bellis, Associate Professor of Empirical Research Methods at the University of St. Gallen.

They conducted four studies about perceptions of VTL, including one field study about ad performance. This final study is of most interest to us as marketers.

The researchers created four distinct ad campaigns that ran simultaneously for 3 days on YouTube’s advertising platform

It targeted exclusively English-speaking users

They used skippable in-stream video ads with a cost-per-impression bidding strategy

35,430 peoples were exposed to the ads

The demographics across all campaigns were approximately 50% between 18–34 years old, 65% of viewers were male, and 35% female.
An 18-second video depicting a static image of a burger with a voiceover promoting a fictional burger brand

The ads were identical except for replacing the word “beef” with “vegan” in two conditions

Each message was communicated by either the low or high VTL VO. 

The main dependent variables were users’ click-through rate (i.e., number of ad clickers divided by number of viewers) and the cost per impression as calculated by the Google Ads platform.

Findings

The use of a congruent voice (longer VTL) in advertising a beef burger led to a higher click-through rate (2.72%) compared to a shorter VTL voice (1.79%)

For the vegan burger advertisement, the shorter VTL voice led to a slightly higher click-through rate (2.18%) compared to the longer VTL voice (2.03%). However, this difference wasn’t statistically significant

Using a congruent voice in the beef burger ad reduced costs by 28.17% and in the vegan burger ad by 6.45%, demonstrating the cost-effectiveness of matching voice characteristics to the product advertised.

Limitations

A significantly higher portion of the total ad viewership was male than female (65% vs 35%), which could magnify the efficacy of masculine-congruent ad sets

Very limited in the scope of products tested