The columnist still wrote the piece. The opinion is still theirs.
In Brazil, one of the country's largest digital media operations has begun lending its columnists a second voice — one shaped by artificial intelligence. UOL is converting written opinion pieces into audio narrations that approximate each author's distinctive cadence and tone, reaching listeners who commute, who cannot read, or who simply prefer to hear the news. The experiment asks a quiet but consequential question: when a voice is synthetic but the thought is human, what exactly has been preserved — and what has been lost?
- UOL is deploying AI voice synthesis not as a gimmick but as an accessibility strategy, targeting audiences who consume news through their ears rather than their eyes.
- The tension is immediate: listeners may hear something that sounds like a familiar columnist without knowing the voice itself was never recorded by a human being.
- Questions of consent, authenticity, and labor hang over the initiative — does this technology quietly displace voice actors, and do readers deserve to know what they are actually hearing?
- UOL's answer, for now, is transparency through design: the columnist still wrote the piece, the opinion is still theirs, and the AI is framed as an access tool rather than a replacement.
- The industry is watching — if audiences accept AI-mediated editorial voices, the model scales automatically, and every new column becomes an audio product the moment it is published.
UOL, one of Brazil's largest digital media operations, has begun using artificial intelligence to convert its written opinion columns into audio narrations — with synthetic voices modeled after the distinctive tones and cadences of the columnists themselves. The move reflects a broader shift in how newsrooms are thinking about AI: not merely as a tool for automating routine tasks, but as a way to extend the reach of their most distinctive editorial voices.
The practical logic is straightforward. Modern audiences consume media in fragmented, mobile ways. Commuters listen. People with visual impairments rely on audio. By generating narrations that carry something of each writer's personality, UOL can serve these audiences without asking already-stretched columnists to sit down and record themselves. Once a voice profile is established, every new column can be converted automatically.
What distinguishes this deployment is its specificity. Rather than applying a generic text-to-speech voice across all content, the system generates audio that reflects each individual author's vocal characteristics — meaning a reader familiar with a particular columnist might recognize something of their rhythm and tone, even knowing the voice is synthetic.
The initiative sits at the intersection of several unresolved tensions. There is the question of consent and transparency — do audiences need to know they are hearing an AI approximation rather than the author's actual voice? There is the question of labor — does this eventually reduce demand for human narration talent? And there is the deeper question of what editorial identity means when AI handles not just distribution but the presentation of voice itself.
UOL's experiment is ultimately a test of audience tolerance and trust. If synthetic voices grow indistinguishable from real ones, the ethical stakes will deepen considerably. If they remain obviously artificial, listeners may find them alienating. For now, the publication is betting that the convenience of audio access outweighs whatever loss of connection audiences might perceive — and that a voice can be synthetic while the thought behind it remains entirely human.
UOL, one of Brazil's largest digital media operations, has begun using artificial intelligence to transform its written opinion columns into audio content, with the synthetic voices modeled after the distinctive patterns and tones of the columnists themselves. The technology allows readers to hear pieces in what approximates each author's own voice—a shift that reflects how newsrooms are experimenting with AI not simply to automate routine tasks, but to extend the reach and accessibility of their most distinctive editorial voices.
The move addresses a practical reality of modern media consumption. Not everyone reads the same way anymore. Some people listen during commutes, while others have visual impairments that make audio the only viable format. By converting columns into audio narrations that carry something of each writer's personality, UOL is attempting to serve these audiences without requiring columnists to sit down and record themselves—a time-intensive process that most opinion writers, already stretched thin, would resist.
What makes this deployment noteworthy is the specificity of the approach. Rather than applying a generic text-to-speech voice to all content, UOL's system generates audio that reflects the vocal characteristics associated with each individual author. This means a reader familiar with a particular columnist's work might recognize something of their cadence and tone in the audio version, even though the voice itself is synthetic. The technology learns from samples or profiles of each writer's speaking style and applies those patterns to the written text.
The initiative sits at an intersection of several ongoing tensions in digital journalism. There is the question of consent and authenticity—do readers need to know they are hearing an AI approximation rather than the author's actual voice? There is the question of labor and efficiency—does this technology eventually reduce the need for human narration or voice talent? And there is the broader question of what journalism becomes when AI handles not just distribution but the presentation of editorial voice itself.
For UOL, the practical benefits are clear. Audio content expands the addressable audience for opinion pieces that might otherwise reach only people sitting down to read. It creates a new product line from existing editorial assets. And it does so without requiring additional time from writers or the hiring of voice actors. The technology, in theory, scales: once a columnist's voice profile is established, every new column can be converted automatically.
The adoption also signals where the media industry is heading. Major publishers are experimenting with AI-generated content in various forms—from automated sports recaps to summarization tools. What distinguishes UOL's approach is that it is not trying to hide the technology or pass off the synthetic voice as human. Instead, it is using AI as a tool to expand access to human-created editorial work. The columnist still wrote the piece. The opinion is still theirs. The voice is the only thing that is synthetic.
Whether this model becomes standard or remains a niche experiment will depend partly on how audiences respond and partly on how the technology evolves. If synthetic voices become indistinguishable from real ones, the ethical questions will only deepen. If they remain obviously artificial, readers may find them jarring or prefer the original text. UOL's experiment is, in effect, a test of whether audiences will accept AI-mediated versions of journalistic voices—and whether the convenience of audio access outweighs whatever loss of authenticity or connection they might perceive.
A Conversa do Hearth Outra perspectiva sobre a história
Why does a news organization need to synthesize a columnist's voice? Why not just have them record it themselves?
Because most columnists are already writing on deadline. Recording, editing, and producing audio takes time they don't have. This automates that step and still gets the piece into audio format.
But if the voice isn't real, doesn't that feel like a kind of deception to the listener?
That's the real question, isn't it. UOL isn't hiding that it's synthetic—at least not in principle. But most listeners won't know the difference, and that's where the ethical line gets blurry.
What happens to voice actors or narrators in this scenario?
That's the labor question nobody wants to answer yet. If every publication starts doing this, there's less work for professional narrators. But right now, it's still niche enough that it's more about expanding access than replacing people.
Does the synthetic voice actually sound like the real author?
It's modeled on their speaking patterns and tone, so there's something recognizable there. But it's not a perfect replica. Listeners familiar with the columnist might pick up on something off about it.
What's the real innovation here—the technology or the business model?
Both, but the business model is what matters more. The technology already exists. What's new is a major publisher deciding it's worth doing at scale, treating synthetic voice as a legitimate way to distribute editorial work.