The Accent Question: What Do We Lose When We Sand Down Our Speech?

Like so many modern personal journeys, this one started with a targeted Instagram ad. The woman in the clip cut straight to a quiet truth no one warns new immigrants about: accent discrimination is not some abstract unfairness—it is very, very real. Her own accent carried a faint Eastern European lilt, so subtle I had to rewatch the clip a few times before I even caught it.

The ad promoted BoldVoice, an AI-powered app built for accent modification training. A few quick clicks landed me on its viral “Accent Oracle” feature, which claimed it could pinpoint my native language after just a short speech sample. After I read off a lengthy prepared phrase, the algorithm confidently declared: “Your accent is Korean, my friend.” It felt almost smug in its certainty—but it was also impressive. It got the answer exactly right.

I have lived in the United States for more than a decade, and my English extends far beyond basic fluency. You could even call it hyperfluent: my diction, for example, lands roughly two standard deviations above the national average. But even that has not been enough to give me a fully native-sounding accent. I started learning English just late enough to miss the critical developmental window that lets people acquire a natural, native accent. That small difference has deadly consequences, as human history makes brutally clear. In the Book of Judges, Gileadite forces used the word “shibboleth” to root out fleeing Ephraimites, who could not pronounce the opening “sh” sound and said “sibboleth” instead—marking them for slaughter. Centuries later, in 1937, Dominican dictator Rafael Trujillo ordered the killing of any Haitian who could not correctly pronounce the Spanish word for parsley, perejil, in what became known as the Parsley Massacre.

So when the Accent Oracle scored my speech, the stakes felt surprisingly high. One run gave me an 89% rating, labeling me “Lightly Accented”; another gave me 92%, putting me in the “Native or Near-native” category. That wide swing was unsettling. On a bad day, I thought, that could have been enough to mark me for harm. To improve my odds, I signed up for the app’s free one-week trial.

Accents carry an inherent “medium is the message” weight. How you speak often reveals far more about your origin, class, education, and background than the actual words you choose. In almost every society, mastering the dominant language’s phonetics becomes a valuable form of social capital.

Now, AI has turned its attention to accent modification, just as it has to almost every other part of modern life. Companies like Krisp and Sanas sell real-time accent “neutralization” tools for call center workers, smoothing out a Filipino agent’s speech to sound more familiar and acceptable to customers in places like rural Ohio. Critics of this technology immediately frame it as “digital whitewashing”—a surrender to an imperial, one-size-fits-all version of English. The conversation often leans into racial framing, too, since ads for these services almost always feature people of color, and most outsourced call centers are based in countries like India and the Philippines. But writing this entire practice off entirely is too hasty.

Adjusting your speech to get ahead socially is nothing new. George Bernard Shaw’s Pygmalion (and its musical adaptation My Fair Lady) revolves entirely around Henry Higgins transforming Eliza Doolittle’s working-class Cockney accent to pass as upper class. Even the eminent German philosopher Johann Gottlieb Fichte worked to drop his rural Saxon accent after moving to Jena, worried his rustic speech would lead people to dismiss his work. This is no outdated relic of the past, either. A 2022 British study found that a strict “hierarchy of accent prestige” persists today, and has barely shifted since 1969. A quarter of working adults reported facing some form of accent discrimination on the job, and nearly half said they had been mocked or singled out for their accent in social settings.

When BoldVoice launched, a commenter in a Hacker News thread about the app wrote: “I’d rather strive toward a world where accents matter less than fixing accents people already have.” That is a nice theoretical sentiment, but it ignores the daily reality of immigrants like me—countless Koreans in the U.S. who constantly navigate the awkward, risky phonetic gap between “beach” and “bitch” or “coke” and “cock.” That sort of online comment is classic sanctimonious pablum, the kind of casual moral high ground only native English speakers can claim, or people who deliberately ignore the daily small humiliations non-native speakers face.

What is more, the harshest judgment of accents does not always come from native speakers. A quiet hierarchy of assimilation often plays out in ESL classrooms and immigrant communities, where an accent draws the line between long-settled immigrants and newly arrived arrivals. The claim that “all accents are equal” is linguistically true, but it is completely false sociologically.

So where do accents come from? For one, the sound units (phonemes) that make up one language do not map neatly onto those of a second. A language’s sound inventory can have as few as 11 phonemes, or more than 100—like the Taa language of Southern Africa, famous for its wide array of click consonants. English has roughly 44 phonemes, and standard Korean has a similar 40. But even that overlap does not mean sounds translate easily. Korean speakers often struggle with the “th” sound, which our brains automatically substitute with a “d” or “s” sound. Conversely, native English speakers trip over the Korean vowel eu (으), which has no equivalent in English’s phonetic system. That is part of the reason I shortened my Korean name, Seung Heon, to Sheon—pronounced just like “Sean.”

After its initial assessment, BoldVoice flagged a few “Top Focus Areas” for me: the “th” sound (no surprise there), devoicing final consonants (it noted I turned the “d” at the end of “did” into a “t” sound), and lengthening the “ee” vowel in words like “seat” so it did not bleed into the short “i” sound of “sit.”

So did my accent actually change after using the app? I will never know. I only made it through three lessons. The ignominy of sitting alone at home repeating “think, thought, thirty” into my phone was too absurd to sustain. It reminded me of that cringey evening I spent in front of my bathroom mirror, fumbling to replicate a dance routine I’d seen on YouTube.

Maybe I would have jumped at this tool when I was first learning English, back when I was deeply self-conscious about my much thicker accent. But as my accent softened over the years, it solidified into something else: a simple auditory shorthand that signals my identity. Working through the app’s exercises made me realize exactly what I stood to lose: if I sanded down every last foreign edge of my speech, I would erase the unique vocal fingerprint that makes me recognizably me.

As it turns out, the woman in that original Instagram ad is one of BoldVoice’s cofounders, an Albanian immigrant. To her, and anyone who finds real use in this app, I say Godspeed. After all, we never know when someone will ask us to pronounce “parsley.”