AI is making everyone sound the same

Something strange has been happening to writing on the internet. Open LinkedIn. Scroll through a few posts. Notice how they all sound oddly similar. Same sentence rhythms. Same vocabulary. Same paragraph structure. Same way of opening with a one-line hook, then dropping to a new paragraph.

You are not imagining it.

The research is in

A study published in Trends in Cognitive Sciences in March 2026 by researchers at USC laid it out plainly. The team, led by Morteza Dehghani, analyzed language across Reddit posts, scientific papers, and American community newspapers from 2018 to 2025. They found a measurable spike in AI-generated text starting in late 2022, right when ChatGPT launched. And alongside that spike, a drop in the variance and complexity of written text.

The writing got smoother. And less diverse.

A separate peer-reviewed paper presented at ICLR 2024, titled "Does Writing with Language Models Reduce Content Diversity?", tested this directly. Researchers had people write essays with and without AI assistance. The group using InstructGPT produced text with measurably lower lexical diversity. Their word choices converged. Their key points converged. The AI was not just helping them write. It was pulling them toward the same center.

An independent analysis of 500 AI-generated LinkedIn posts, published on DEV Community, found that 82% used identical opening structures. 91% followed the same formatting pattern. The vocabulary overlap was staggering. Five hundred posts, theoretically from five hundred different professionals, reading like variations on the same template.

And Originality.ai estimated that 54% of long-form LinkedIn posts are now AI-generated. More than half.

How this happens

The mechanism is straightforward. Large language models are trained on averaged language. They learn patterns from billions of documents and produce output that reflects the statistical center of all that text. When a thousand people use the same model to draft their LinkedIn posts, they get output drawn from the same distribution. The edges get smoothed. The distinctive choices get replaced with probable ones.

"Interesting" becomes "fascinating." "I think" becomes "I believe." "Let me explain" becomes "Here's the thing." The model's vocabulary preferences override the writer's vocabulary preferences, and since every writer is using the same model, everyone ends up in the same voice.

USC's research called this "cultural homogenization." Their paper warned that LLMs disproportionately reflect a narrow demographic: western, liberal, high-income, highly educated, English-speaking. When these models mediate communication for everyone, the output converges toward that single register. Alternative voices, styles, and reasoning strategies get marginalized.

It is not about quality

The homogenized writing is not bad writing, by conventional standards. It is grammatically correct. It flows well. It is clear. By most readability metrics, it scores fine.

That is exactly the problem. Quality is not the issue. Identity is.

When your professional communication sounds indistinguishable from the next person's, you lose something harder to measure than readability. You lose recognizability. Your colleagues stop associating your emails with your specific way of thinking. Your readers stop feeling like they are hearing from a person.

The research paper from ICLR puts a number on this. They measured "content diversity" across groups of writers. The groups using AI produced writing that was more similar to each other. Not similar in quality. Similar in substance and style. The AI collapsed individual differences.

The LinkedIn problem is a preview

LinkedIn is where this is most visible, but it is happening everywhere. Marketing copy. Internal memos. Newsletter drafts. Student essays. Blog posts. Customer emails. Wherever AI assists the writing process, the output trends toward the same voice.

And the feedback loop accelerates it. People read AI-generated content. Their sense of "normal" writing shifts toward the AI's register. When they write without AI, they unconsciously mimic the patterns they have been reading. The model influences the data that trains the next model. The center keeps tightening.

Dehghani's team warned about exactly this: when distinct linguistic styles and reasoning strategies are mediated by the same LLMs, they become homogenized. Standardized expressions replace individual ones. And over time, people stop noticing the difference because the new normal is all they see.

What you can do about it

The fix is not to stop using AI. The fix is to stop letting AI replace your voice with its default.

That starts with knowing what your voice sounds like. Most writers have never studied their own patterns. They have never measured their sentence length distribution or cataloged their vocabulary habits. They write intuitively and assume the result is "their voice." But when AI starts inserting its patterns into their drafts, the intuition breaks down.

Here is a practical test: take something you wrote by hand six months ago, before you started using AI for drafts. Put it next to something you wrote this week with AI assistance. Read both out loud. The differences are usually obvious once you look for them. Sentence length gets more uniform. Vocabulary gets flatter. Opening structures follow a pattern.

The second step is anchoring your writing to your actual patterns. If you have a reference, a set of samples that represent how you genuinely write, you can course-correct when a draft drifts toward the generic center.

Yourtone was built for exactly this. You feed it samples of your real writing, and it extracts the patterns that make you sound like you. When you paste text for rewriting, it applies your patterns, not the model's default. The output belongs in the same family as your actual writing, not in the same family as everyone else's AI drafts.

The value of sounding like yourself

There is something ironic about using AI to sound more like yourself. But that is where we are. The default output of every AI writing tool is the same voice. If you want to sound different, if you want to sound like you, you have to be deliberate about it.

The alternative is letting your writing blend into the noise. And there is a lot of noise.