Free tools Windows power users keep installed
One-click scans. No signup required.
Neither AI nor human writers are best at every kind of writing. In studies of argumentative essays, short fiction and business refusal messages, results changed with the task and the way quality was judged. AI scored well on one essay rubric; human-written refusal messages scored highest overall; and readers of one fiction experiment reported less immersion in AI stories. The practical choice depends on whether you need structure, nuance, reader engagement or a range of distinct ideas.
What does “better writing” mean?
A writing score, a reader’s experience and the result of using AI as an assistant are different outcomes. A text can score highly for organization and language while still feeling less personal to a reader. Likewise, an AI-assisted draft may help one person develop an idea without making a collection of writers’ work more varied.
As an Amazon Associate I earn from qualifying purchases.
That distinction matters when interpreting the studies below. They examine different genres, participants, AI systems and measures—not a single contest in which every kind of writing is ranked on the same scale.
How did AI and human writing compare in these studies?
| Task and study | What was evaluated | Finding | What the result applies to |
|---|---|---|---|
| Argumentative essays — Herbold, Hautli-Janisz, Heuer and colleagues, Scientific Reports, published October 30, 2023 | Teachers scored essays by German high-school students from an online forum and essays generated by ChatGPT versions available at the time, using content and language criteria. | The ChatGPT essays scored higher than the sampled forum essays across the study’s rubric criteria. ChatGPT-4 also scored above the earlier ChatGPT version on measures including logical structure, language complexity, vocabulary richness and text linking. | Student argumentative essays, those model versions and that scoring rubric—not writing quality across all genres or professional writers generally. |
| Short fiction — Appel, Malecki, Messingschlager and colleagues, Humanities and Social Sciences Communications, 2025 | One hundred students and ChatGPT wrote stories from similar prompts. The study compared 200 stories in total; 380 participants read stories from the pool and reported their reactions. | Readers reported no difference in novelty or entertainment, but reported less narrative transportation—the feeling of being drawn into a story world—for AI-generated stories. Linguistic analysis found differences in pronouns, descriptions of relationships and positive emotion language. | The prompts, student writers, ChatGPT and reader experiment in this study; it does not settle how readers respond to every story or professional fiction. |
| Business refusal messages — Wilson and Rose, Business and Professional Communication Quarterly, first published online March 12, 2025 | Assessors evaluated human-written messages and messages generated by ChatGPT and Gemini, considering features such as tone, relationship, language, content and structure. | Human writing received the highest quality ratings overall. Gemini’s ratings were not significantly different from the human scores in this sample; ChatGPT scored lower. The authors characterized AI messages as more formulaic and less nuanced. | An exploratory comparison of refusal messages and the systems tested. A non-significant Gemini-human difference does not prove the two perform equivalently in general. |
These results cannot be combined into a universal score for “AI writing.” The essay study used teacher rubric scores, the fiction study measured language and reader-reported reactions, and the business study assessed a specific workplace genre. Each can answer a different question about writing.
#1 Best Overall
What can AI writers do better?
Produce rubric-friendly argumentative essays
In the 2023 essay study, the tested ChatGPT essays outscored the sampled high-school forum essays on the teachers’ content and language rubric. That is evidence that those models could produce strong work under those conditions, especially on criteria such as structure and language. It is not evidence that AI is a better writer than an expert in every context, or that a high rubric score guarantees original insight or a persuasive effect on every reader.
Help an individual develop ideas
A 2024 Science Advances experiment found that generative AI assistance enhanced individual creativity among non-professional writers using GPT-4. In that experiment, however, the collective diversity of novel content fell. The two outcomes are compatible: assistance can improve an individual participant’s result while nudging a group’s output toward greater similarity. The finding is limited to the experiment; it does not quantify the effect for all writers or professional teams.
Where did human writing have an advantage?
Nuance in relationship-sensitive messages
A refusal is not just a statement of “no.” It also has to manage tone and the relationship with the recipient. In Wilson and Rose’s study, human-written refusal messages scored highest overall, and the AI texts were described as more formulaic and less nuanced. That makes the finding most useful for writing where the manner of refusal matters, rather than as a general claim that humans outperform AI at workplace writing of every kind.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minuteImmersion in the tested fiction experiment
In Appel and colleagues’ 2025 experiment, readers reported feeling less transported into AI-generated stories, even though they gave AI and human stories similar ratings for novelty and entertainment. Those measures should not be collapsed into one judgment: readers can find stories entertaining without feeling as absorbed in their worlds. The study does not establish that AI fiction is always less immersive.
Should you use AI, a human writer or both?
- For a structured first draft or an argument judged against explicit criteria: AI may be useful for generating an outline or draft to review. The essay evidence supports only a bounded claim about the tested models, student essays and rubric.
- For a sensitive refusal or message where preserving trust is central: have a person shape and review the wording. The business-message study points to nuance and relationship management as important differences in that genre.
- For fiction where immersion is the goal: treat AI-generated prose as material to evaluate and revise, not as a guaranteed substitute for a writer’s choices. The reader study found similar novelty and entertainment ratings but lower reported transportation for AI stories in its sample.
- For a team seeking a distinctive range of ideas: consider whether shared AI assistance could make outputs more alike. The creativity experiment found higher individual creativity alongside lower collective diversity among its participants.
A practical workflow is to give AI bounded, reviewable tasks—such as proposing an outline, listing counterarguments or suggesting alternate phrasings—then have a writer check the claims, decide what the piece should say and adapt its voice to the audience. This is a decision framework, not a measured guarantee that such a workflow will improve every text.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What can’t these findings tell you?
The studies do not provide a single comparable statistic ranking AI and humans across genres. They also do not establish that human writing is always more accurate, that AI is always faster, or how every writing system available in October 2026 performs. The evidence concerns the prompts, populations, genres and systems each study actually tested. A result about student essays should not be transferred automatically to fiction or business communication, and a finding about AI assistance is not the same as a head-to-head comparison of finished texts.
Quick Recap
Best Value
Rank #4
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
