Grammarly plans to introduce an AI feature that learns a user’s writing style and can apply it to other text. The tool is intended to help people produce writing that sounds more personal, but it also raises questions about who controls a writer’s digital voice and how that voice may be used.
How Grammarly’s voice profiles work
The feature, called “Personalized voice detection and application,” was scheduled to launch by the end of the year for subscribers to Grammarly’s business tier. It builds a profile by observing how a person writes while using Grammarly across apps and tools.
Grammarly says the profile includes an AI-generated description of the user’s style, such as whether their writing tends to sound “positive” or “encouraging.” Users can adjust the profile, including removing tone or style preferences they think do not reflect how they write. The examples given include active voice and compound sentences.
Tal Oppenheimer, Grammarly’s head of product for client apps, said the profile is created as people use the product. She described the feature as an early release that Grammarly expects to refine over time.
The company presents the feature as a way to use context to make writing sound more personal. A profile could help a user shape new text to fit their usual voice without having to describe their preferences from scratch.
Questions about consent and impersonation
The same ability to reproduce a writing style can raise concerns beyond convenience. If an organization could use a writer’s profile without their approval, it might publish work that sounds like the writer’s even when they did not create or authorize it. A convincing imitation could also be used to impersonate someone in a phishing attempt.
The article points to a recent example of fake work appearing under a writer’s name. In August, author Jane Friedman found books for sale on Amazon under her name that she said she had not written and that appeared to have been generated by AI. Amazon later removed the books and said its policies prohibit such imitation.
That episode illustrates why a writing style can matter as more than a matter of preference. A recognizable voice may be closely associated with a person’s work, and AI tools that reproduce it prompt practical questions about attribution, permission and compensation.
Limits at launch, uncertainty ahead
Grammarly said that, at launch, an individual user can access only their own voice profile. Profiles cannot be exported. Those limits constrain how the feature can be used initially, but they do not settle what controls may exist as the product develops.
The concern raised in the article is whether a business using Grammarly might eventually seek access to the profiles of its writers, potentially to create content when those writers are unavailable or no longer employed. Grammarly’s marketing emphasizes cost savings, but the article does not establish that the company plans to give businesses this kind of access.
Oppenheimer said the profiles are not designed to replace anyone. She described them as a way for writers to understand how they sound and create more personalized, authentic output. The feature’s stated purpose and its potential uses are therefore part of an ongoing debate about how much control a person has over an AI-generated version of their voice.
A wider debate over AI and writers
Concerns about imitation extend beyond Grammarly. The article notes that thousands of authors signed an open letter criticizing generative AI technologies for mimicking and regurgitating their language, stories, style and ideas. Writers in California and New York have also sued AI startup OpenAI, alleging that it trained text-generating AI on their work without permission.
Grammarly’s planned tool brings that debate into everyday writing software. The company said it had more than 30 million users and over 70,000 teams on its platforms, and that users were creating over 12 million pieces of content weekly with its generative AI features. As tools that adapt writing style become more common, the questions raised by voice profiles will turn on how access is controlled and whether writers have a say in how their work and identity are represented.