Why AI Writing Tools Struggle to Capture a Person’s Voice

In a recent LinkedIn post, Justin Oberman discusses why AI writing tools that promise to write in a person’s voice can fall short. This article covers Oberman’s experience and analysis; it is journalistic reporting about his post, not an article written by him.

“I spent this whole week waiting for TypeShare to write my LinkedIn posts.”

Oberman describes testing TypeShare’s new “digital brain” as part of his work reviewing AI writing tools for Copy Hunter Lodge members. After setting up a LinkedIn agent and waiting a week, he received five posts that, as he puts it, he had already written. When he rejected them and requested another set, the results had the same problem.

When generated posts echo the source material

Reading the second batch together, Oberman recognized familiar material in different forms. He says the first post resembled one of his own with altered verbs, the second was a weaker version, and the remaining three applied one formula to different topics.

“It either just repeats posts and topics that performed well or it writes everything using the same formula but decorated with the words and phrases you use.”

Oberman cautions that his complaint is not limited to TypeShare. In his account, the broader issue is that some tools reuse successful past content, apply a repetitive template, or combine both approaches while presenting the result as a writer’s distinctive voice.

Why “perfect” imitation may be the wrong goal

Oberman connects his observations to his background in the philosophy of language. He says that when he does use AI for writing, he combines projects and skills he has trained to help capture his style. Even with that approach, he estimates the result gets 75% to 90% of the way there, leaving a human to make the remaining edits.

“Actually, the concept of perfect when it comes to language is part of the problem.”

That point frames Oberman’s critique: language is not simply a collection of recurring phrases or a template that can be copied exactly. In his view, tools that focus on surface patterns risk reproducing familiar wording without capturing what makes an individual’s writing distinctive.

A promised alternative

Oberman says he plans to explain what he believes a writer’s voice and style involve, why many AI cloning tools underperform, and what users can provide to AI instead. He announced a free webinar for the following Wednesday, November 11. His post does not detail the full method, but it presents the event as a response to what he sees as misleading claims about automated voice imitation.

For readers evaluating AI writing products, Oberman’s account offers a practical test: look closely at whether a tool produces new work that reflects a writer’s approach, or mainly reshuffles old posts and applies a recurring formula. His conclusion is that AI can help with a draft, but a person still needs to shape it before it is ready.