‘Shark Tank’ Study Shows That Your Voice May Be Key To Landing Deals, Persuading Investors

A voice that resembles a listener’s own may be working on that listener’s wallet. New research finds that when a spokesperson, pitchman, or recommender sounds vocally similar to the person hearing them, that person trusts them more and is more likely to be persuaded. The similarity in question is not accent or pitch, but timbre, the particular texture and color of a voice that makes one person’s “hello” sound different from another’s.

Researchers at the University of Cincinnati, Georgia Institute of Technology, and the University of Michigan ran six studies measuring this effect, and the pattern held up across pitch contests, crowdfunding campaigns, radio ads, restaurant picks, headphones, and seltzer. The study found that people leaned toward the voice that sounded more like their own. They believed that voice was more competent, and they followed its advice more often.

Timing gives the work a sharper edge. Voice-cloning tools can now copy a person’s voice from under a minute of audio, and AI assistants such as Siri and Alexa are everywhere. A finding about whose voice people trust lands differently in a world where a company can build a voice to order.

Measuring vocal similarity by the numbers

Past studies of “similarity” usually leaned on human judges, who would listen to two voices and rate how alike they sounded. That approach is slow and subjective. So the team built an objective yardstick instead, using a tool from audio engineering called mel-frequency cepstral coefficients, or MFCCs. In plain terms, MFCCs break a voice into a set of numbers that capture its timbre, qualities that tend to stay fairly stable for a given speaker even as the words change. A smaller mathematical distance between two voices means they are more similar. As the authors put it, their method “utilizes machine learning methods to capture similarity in a manner beyond human hearing.”

That measure became the engine for everything that followed. Two voices could be compared without anyone having to guess.

Vocal similarity on the biggest stages

To test the idea where real money was on the line, the researchers started with “Shark Tank.” They analyzed 7,002 entrepreneur-investor pairings from the show and checked whether vocal similarity between a pitcher and a shark predicted a deal. It did. Greater similarity in timbre between an entrepreneur and an investor was linked to a higher chance of that pair closing a deal. Similarity in vocal pitch, by contrast, did not predict success, and neither did a gender match between the two. Timbre was doing the work.

Crowdfunding came next. Across 2,091 Kickstarter campaigns archived over more than a decade, the team asked a slightly different question: what happens when a spokesperson is talking not to one person but to a crowd? Here they compared each campaign narrator’s voice to the average voice of the whole sample, a stand-in for the “average” listener. Campaigns whose narrator sounded closer to that average voice raised more money and were more likely to hit their goal. A voice near the middle of the pack reaches more people, because more people hear something familiar in it.

One detail complicates the neat story, in a useful way. When a campaign carried an outside stamp of credibility, a Kickstarter “staff pick” badge, the voice mattered less. Given a strong outside signal, listeners leaned on that instead of on how the narrator sounded. Vocal similarity, then, is one cue among many, a thumb on the scale rather than the whole scale, and it loses weight when better information shows up.

Trust earned through a familiar-sounding voice

Field data can show a pattern but cannot fully rule out coincidence, so the team ran four controlled experiments. In one, 155 participants listened to five radio ads, then recorded their own voices so the researchers could compute similarity for each person individually. Listeners rated the more vocally similar spokesperson as more persuasive and reported being more likely to buy. The result held even after accounting for the participants’ own vocal pitch.

In another study, 185 people chose among three pizza restaurants recommended by three different voices, matched to each listener’s gender. People tended to pick the restaurant pushed by the voice that sounded most like theirs. Later experiments raised the stakes with real prizes on offer for headphones and a sparkling drink, and added a twist: participants rated how similar each voice felt to their own. Subjective similarity, the sense that a voice resembles one’s own, produced the same pull as the objective measure. What people felt about a voice tracked closely with what the math said about it.

A consistent thread ran through the experiments. A similar voice raised trust specifically in the speaker’s competence, the sense that this person knows what they are talking about, and the studies point to that trust as a key pathway through which persuasion rose. Familiarity dulled the effect slightly. With a product people already knew, a similar voice still made the recommender seem more persuasive, but it had less sway over whether someone wanted to try the product.

Why a familiar voice wins people over

Two careful explanations come from the researchers, both labeled as possibilities rather than settled fact. One is cognitive fluency: a voice that resembles a listener’s own is easier for the brain to process, and ease tends to feel like trustworthiness. The other draws on an old idea in psychology called balance theory, which holds that people use cues around them to make sense of the world. A familiar-sounding voice may signal that the speaker shares a listener’s traits, and therefore understands what that person wants. Both ideas need more testing, and the authors say so plainly.

Practical reach is where the finding gets uncomfortable. For now, a company can only aim for the broad middle, choosing voices close to an average listener. But the authors sketch a nearer future in clear terms: a call center could capture a customer’s voice in real time and adjust an AI agent’s voice to sound more like that customer’s, one person at a time. With cloning tools that need less than a minute of audio, the technical barrier is already low.

That possibility puts a familiar marketing trick on new footing. Brands have long picked spokespeople for their voices. A system that reshapes a voice to match each individual listener is a different thing, and the researchers do not pretend otherwise. They flag it as an open question for both science and policy. “It is an important question for future research to study in which contexts this would be considered helpful for the consumer and ethical for a firm to do,” they write. Trusting a voice that sounds like one’s own is human. Knowing a machine built that voice on purpose is the part worth thinking about.

---