Is AI making us all write the same way?

Artificial intelligence writing assistants can correct, summarize and rephrase in seconds. That efficiency is real. It also comes with a less visible effect: when many people use the same tools in the same way, their writing begins to look more alike.

A recent large-scale analysis of more than 880,000 texts from several domains observed a decline in linguistic diversity as writing assistants became more common. In rewriting experiments, the central meaning was generally preserved, but variation in complexity and style fell by 21 to 50 percent depending on the dataset and model.

The meaning remains while the voice narrows

Two people can express the same idea with different rhythms, words and structures. Those choices are not merely imperfections. They carry traces of experience, culture, profession, region and sometimes a person’s state of mind.

When an assistant receives a broad instruction such as “improve this text”, it tends to return a fluent, orderly and cautious standard version. The result often looks cleaner. Personal turns of phrase, useful hesitation, local references and distinctive humour may disappear. Each text still works on its own, while the collection becomes more uniform.

Writing alike is not the same as thinking alike

An exaggerated conclusion would be misleading. Convergence in writing style does not prove that everyone now thinks in the same way. It shows that the same statistical intermediary can influence how different people present their ideas.

The risk becomes greater when the tool is used before the first independent idea. People who ask similar questions often receive related suggestions arranged in comparable structures. AI can then influence the starting point, the vocabulary of the discussion and the options that initially appear legitimate.

Why diversity has practical value

Linguistic diversity is not just aesthetic. It helps us understand how different groups describe a problem, recognize a danger or express a need. Excessive uniformity can impoverish school work, professional communication, cultural narratives and public consultation.

In sensitive settings, language is sometimes used as an indicator of understanding, development or possible difficulty. Automated rewriting can hide those signals. A more standardized sentence therefore guarantees neither more accurate thinking nor a fairer decision.

Using AI without losing your voice

A few straightforward habits can preserve critical distance:

• write your ideas, examples and conclusion before asking for assistance;
• request targeted criticism instead of a complete rewrite;
• ask for several conflicting alternatives rather than accepting the first answer;
• retain local expressions, lived examples and professional vocabulary;
• compare the suggestion with the draft and restore deliberate personal choices;
• read the final version aloud to check that it still sounds like its author.

Teams can also adopt a clear rule: AI may help correct, translate or structure a document, but the facts, viewpoint and final wording remain a human responsibility.

A dialogue tool, not a mould

The answer is not to reject artificial intelligence. It can make writing more accessible, help overcome a blank page and improve comprehension. The problem begins when an average suggestion automatically becomes the norm.

The most useful approach preserves the effort that produces thought: selecting facts, forming a position, introducing nuance and deciding what deserves to be said. The tool can support that work. It should not erase the differences that make a human voice recognizable.