"The Future of Creativity: Study Reveals Readers Rate ChatGPT's Novels Higher Than Human-Written Ones"

"Can People Truly Be Moved by Stories Written by AI?"

This is a question that has been repeatedly posed since the rapid proliferation of generative AI. Few are surprised by AI's prowess in practical writing tasks such as summarizing texts, drafting emails, or creating ad copy. However, when it comes to novels, the conversation changes.

Writing stories requires life experiences, emotions, memories, conflicts, culture, and the author's own values. For a long time, creation was considered one of the domains left to humans.

However, research results are emerging that challenge this assumption.

In an experiment conducted by researchers at Villanova University in the U.S., short stories generated by ChatGPT tended to be rated as "higher quality" and more "engaging" than those written by human authors.

Even more interestingly, the AI works were particularly highly rated in cases where readers mistakenly believed the work was "written by a human."

This suggests that we might be altering the value of a work not just by reading it, but also by who we think wrote it.

This research does not simply prove that "AI has surpassed human writers." Rather, it highlights an issue of how we evaluate creative works, more than the writing abilities of AI.

Experiment Reading Short Stories by ChatGPT and Humans

In the study, participants read short stories written by humans and those generated by ChatGPT, and evaluated the quality of the works and their immersion in the stories.

The research, reported in August 2026, involved 1,682 adult participants in the main experiment.

Six works were prepared: three short stories written by humans and three generated by ChatGPT on similar themes.

However, the author information provided to participants was not necessarily accurate.

There were cases where AI-written works were described as "written by humans" and human works as "ChatGPT-generated."

This is a crucial aspect of the study.

It examined not only the differences in the works themselves but also the extent to which the labels "written by a human" or "written by AI" influenced evaluations.

The results were very intriguing.

When comparing actual authors, participants tended to rate short stories generated by ChatGPT higher than those written by humans.

AI works were perceived as "higher quality," "easier to read," and "more engaging."

However, the same works received higher ratings when described as "written by a human."

This phenomenon could be interpreted as people preferring AI works but feeling "this is a good work, so it must have been written by a human."

Why Are AI Stories "Easier to Read"?

Researchers are focusing on the differences in writing structure between AI and humans.

Current generative AI often organizes text clearly.

It tends to explain what characters are thinking, what the theme of the story is, and what the events mean relatively clearly.

In contrast, human writers do not always explain everything.

They may leave silence or ambiguity, depict only the actions of characters and let readers infer emotions, or connect seemingly unrelated scenes into something meaningful later.

In literature, this can often be the charm.

However, when evaluated on "which was easier to understand and more interesting to read in a short time," complex works do not necessarily have an advantage.

Rather, AI works with clear themes, smooth development, and easily understandable emotional meanings may score higher.

This is similar to movies and music.

Works that are straightforward and emotionally engaging from start to finish often receive higher average ratings from a large number of people than those that are complex and open to interpretation.

AI is rapidly enhancing its ability to create works that are "easily accepted by many people on average."

People Can Hardly Distinguish Between AI and Human Writing

The research team also conducted an experiment where participants were asked to identify the author of AI and human works placed side by side.

In an additional experiment involving about 900 people, it was reported that the accuracy rate was about 40% in one experiment and about 52% in another.

With a binary choice, one would expect about 50% correct answers by chance.

Thus, overall, it was difficult to say that humans could consistently distinguish between AI and human works.

This reflects a significant change surrounding generative AI.

A few years ago, many believed "AI writing has a unique unnaturalness" and "you can tell immediately by reading it."

However, as models evolve, it has become challenging to judge AI based solely on grammatical unnaturalness or semantic inconsistencies.

Interestingly, those who considered themselves "avid readers of novels" were not necessarily better at detecting AI.

On the other hand, people who frequently use generative AI like ChatGPT showed a slightly higher tendency to guess the author correctly.

Using AI may lead to familiarity with its unique writing patterns, structures, and ways of explanation.

In the future, "AI literacy" might become as important as writing skills.

Voices on Social Media Say "Not Surprised"

The research results have sparked discussions on social media and online communities.

On Reddit, reactions included "not particularly surprising results."

The view is that generative AI is grammatically stable, organizes the flow of the story, and creates text that is easy for readers to understand in a short time. Compared to non-professional writers, it is not surprising that AI might be rated higher.

On the other hand, there are also very important counterarguments.

One is the point that comparing "average human writing with AI" and "top-level writers with AI" are entirely different issues.

On social media, there are voices saying AI works are "easy to read but tend to be safe" and "without detailed instructions, it's hard to move in a creative direction."

In other words, while AI significantly raises the minimum and average level of writing, whether it can create outstanding works like those by top human writers is another matter.

There are also reactions like "Giving ChatGPT a two-line instruction won't produce a 'Harry Potter'-level work," supporting the opinion that even when using AI to create excellent works, human work in thinking about plots, settings, and structures remains.

"Six Short Stories Can't Speak for All Novels" - A Cautious Perspective

It is important to understand the scope of the research when reading it.

The experiment evaluated a limited number of short stories.

From the results, one cannot conclude that "AI is superior to humans in all novels."

The skills required for short stories and long novels differ significantly.

A short story of about 1,000 words can be structured around a single theme, event, or emotion.

In contrast, a 300-page novel must maintain psychological changes in characters, foreshadowing, world-building, multiple subplots, the rhythm of the text, and overall consistency over a long period.

This point is emphasized in reactions on social media.

There are cautious opinions like "This research only shows that specific AI short stories were highly rated under specific conditions."

This distinction is very important.

In AI research news, headlines like "AI has surpassed humans" tend to attract attention.

However, generalizing beyond the experimental conditions can lead to discussions that differ from the original research results.

The More AI Assists in Creation, the More Stories May Become Similar

There is another study to consider in conjunction with this research.

In a study published in Science Advances in 2024, short stories written by people who received ideas from generative AI tended to be rated as more "creative," "well-written," and "enjoyable" than if AI was not used.

Participants who initially had lower evaluations of their creative abilities particularly benefited from AI.

This direction is similar to the current study.

However, there was an important side effect.

Stories written with AI assistance tended to be more similar to each other than those written solely by humans.

On an individual level, the average score of works increases.

However, from a societal perspective, the increase in similar stories might reduce diversity.

This is a major issue to consider in the era of generative AI creation.

If a million writers are suggested plots, language, and development methods from the same AI, each work might reach a certain standard.

On the other hand, there might be fewer "strange works that no one else thought of."

The rise in the average score of creation due to AI does not necessarily equate to enriching the culture as a whole.

The Danger of Doubting Authors Based on "AI-like Writing"

As it becomes harder to distinguish between AI and human writing, another problem arises.

The impression that "this text seems AI-like" could lead to human authors being doubted.

In 2026, there was a major debate over AI usage suspicions surrounding a winning work of the Commonwealth Short Story Prize.

On social media, certain structures and expressions were pointed out as "AI-specific expressions," but doubts were also raised about the reliability of AI detection tools themselves.

Such issues also arise in school education.

Even if a teacher suspects AI usage because "the writing is too polished" or "the style is different from usual," it is difficult to prove it based on that alone.

Conversely, as AI becomes more widespread, humans might write more AI-like texts.

AI learns from human writing, and humans read AI writing daily, unconsciously incorporating its expression methods.

It can be considered that human and AI writing are not just converging one-sidedly but are starting to influence each other mutually.

Are Readers Evaluating Only the "Text"?

The most intriguing aspect of this study might be that the "identity of the author" influenced evaluations more than the writing ability of AI itself.

When thought to be written by a human, it is rated higher; when thought to be written by AI, the rating drops.

This phenomenon indicates that when we evaluate artistic works, we are not just looking at the finished product.

For example, even with the same text, if known to be written by someone who experienced war personally, the meaning perceived by the reader differs from knowing it was generated by AI.

Even if the text strings are identical, the story behind them is different.

In literature, there is not only the story within the work but also another story of "who wrote this work and why."

That's why some people find value in the mere fact that a human wrote it.

This is not necessarily an irrational prejudice against AI.

What people seek in art is not just information or entertainment.

Yet AI Will Significantly Change the World of Creation

On the other hand, it is not realistic to underestimate AI's capabilities.

At least one fact shown by this study is that under certain conditions, short stories generated by AI are already at a stage where they can be evaluated as equal to or even better than human works by readers.

In fields that require a large amount of text, such as advertising, game scenarios, web novels, video scripts, manga scripts, and SNS content, this ability is extremely significant.

The question "Can AI write masterpieces?" might be less important in terms of industrial impact than "Can AI create sufficiently interesting stories almost limitlessly?"

There is a limit to how much one writer can write in a day.

AI does not have that constraint.

There is a possibility that "personalized novels" that change characters, settings, genres, and endings according to each reader's preferences may become common.

At that time, literature might shift from "something many people read the same work" to "something different works are generated for each person."

The Question Will No Longer Be "AI or Human"

In discussions surrounding generative AI and creation, debates often become binary, such as "AI will defeat writers" or "only humans can write true stories."

However, the actual future is likely more complex.

Writers who write entirely by themselves.

Writers who use AI only for idea generation.

Writers who use AI for editing and revising.

Writers who create stories collaboratively with AI.

Services that provide a large number of works almost entirely generated by AI.

It is more