AI writing tools make everyone's prose look similarly polished — but does that erase real skill gaps, or just hide them from hiring managers?

This explores how writing ability is spread across different groups of people. The collection has no census of who writes well. It has something arguably more useful right now: evidence about what happens to visible differences in writing ability once AI writing tools are common. The short version is that AI shrinks the differences readers can see, but the differences in ability underneath haven't gone away.

The clearest evidence that writing ability varies in a way that matters comes from hiring. In a simulation of Freelancer.com hiring with written signals removed, the strongest workers (top fifth) got hired 19% less often and the weakest (bottom fifth) got hired 14% more often Does cheap writing weaken hiring based on worker ability?. Writing worked as a signal because good writing took effort and skill, and that skill tracked how well people actually did the job. When writing becomes cheap, employers lose a way to tell workers apart, but the workers themselves are still just as different. Evaluation studies point the same way: readers rate AI-polished documents above human ones and often can't tell which is which Does polished writing actually signal better quality work?.

The benefits of AI help aren't spread evenly either. In a controlled experiment with 118 people, GPT-4o autocomplete gave American writers bigger productivity gains than Indian writers, while pulling the Indian essays toward Western phrasing and cultural references Do AI writing assistants push non-Western writers toward Western styles?. Indian participants also accepted more AI suggestions, and the authors argue this reflects cultural patterns of trust in technology, not noise to be controlled away Is higher AI use by Indian writers a confound to control?. So how much AI helps you depends partly on how close your writing culture is to the model's training data. That adds a new kind of unevenness on top of individual skill.

The most surprising finding concerns social signals. In a study of 2,939 writers and 11,091 readers, AI assistance shifted all 29 measured traits of how readers saw the writer, including making writers seem more confident and more privileged Does AI writing assistance change how readers perceive the writer?. Writers edited AI paragraphs only 23% of the time, and their edits left the text 96% the same Do writers actually edit AI-generated text before publishing?, so this effect reaches readers mostly unfiltered. In other words, the writing cues readers once used to guess someone's education or class are being replaced by the model's default voice. That voice is narrow: models cluster tightly in value space while human respondents scatter widely Do large language models actually reflect human value diversity?, and shared reliance on the same models pushes writing toward the same stances and framings Do large language models narrow human expression and thought?.

Homogenized writing also has a cost for readers. In online discussions, AI-assisted comments raised participation but were seen as generic and less authentic, and that drop in perceived quality affected even conversations among people who didn't use the tools Do AI writing tools improve online discussion or degrade it?. If you came looking for baseline data on literacy or writing ability across countries or social classes, this collection doesn't have it. Its contribution is the newer question: what happens to a society's writing when the visible differences get flattened?