When users ask ChatGPT or any AI platform to "write" a few sentences or a long passage on a specific topic, they quickly receive the desired output in seconds. If the tone is not to their liking, they instruct the AI to change and refine every word and detail. Sometimes, the chatbot edits and upgrades drafts into complete versions. Depending on the prompt, artificial intelligence can 'create' a novel or a poem without an outline, as long as the topic is provided.

Does AI-generated writing hold value, or is it merely polished text? Since ChatGPT's launch in late 2022, the literary community has debated this issue. Most oppose AI's involvement, arguing that machine-generated literature is hollow. They contend that large language models (LLMs) only gather human language from online articles and historical literary works. Consequently, AI can quickly mimic the writing styles of many authors but lacks depth.

Logos of several common AI applications. Photo: Bao Lam

According to an article by Zoe Yu, a Harvard University student, in The Boston Globe, the more one reads chatbot-generated content, the more they notice a "peculiar superficiality" in AI's prose. Yu cited an AI passage describing its own writing as "a talking mirror: knowing shape but unable to express soul." The AI then compared its mimicry of "the winding rhythm of a sentence like a dancer learning steps from video, but never waking up with a lump in their throat and deciding to tell the story differently." Yu finds AI's writing awkward, leaving a feeling that "something is off" when read.

On X in 2025, Sam Altman, co-founder of OpenAI, garnered attention when he posted a short story created by the chatgpt-5 model based on the prompt: "Please write a metafictional literary short story about AI and grief." The Guardian reported that the work divided artists, writers, and readers. Some praised the AI's story as "beautiful and moving," while others criticized it as empty.

Inspired by Sam Altman's story, New York Times writer Sam Kriss used ChatGPT to write two stories and an essay. All three AI responses were rich in imagery, emphasizing silence even when the topic involved noise. For example, when writing about a party, the AI began by describing "the gentle murmur of distant conversations" and "the trees outside whispering secrets."

Sam Kriss suggests that AI understands that good writing requires subtlety, expressed by not stating everything directly. To achieve this effect, the chatbot "shouts at full volume" about silence and ambiguity, speaking in half-truths to compel readers to reflect. Coupled with AI's tendency for elaborate writing, Kriss noted that this creates a distinctive chatbot voice: "always a bit surprised, overly enthusiastic, bland yet almost excessive."

Recently, some works faced criticism and were withdrawn from the book market due to suspected AI involvement, such as Mia Ballard's novel *Shy Girl* and *The Serpent in the Grove* by Indian-born writer Jamir Nazir. These works were noted for confusing prose and an overuse of AI-like phrasing and sentence structures.

ChatGPT describes a rainy afternoon scene. Video: ChatGPT

Beyond literature, artificial intelligence has also permeated journalism. In October 2025, the Reuters Institute published the results of a survey on the attitudes of 12,000 people in six countries: Argentina, Denmark, France, Japan, the UK, and the US, regarding AI's role in journalism and society. Of these, 33% of people expressed comfort with AI writing news, and 41% chose to read AI-generated headlines. The most accepted activity (55%) involved AI correcting spelling and grammar for news reports.

In discussions with the institute, journalist Alex Mahadevan stated that what bothers him when reading AI-written news is its tendency to make sweeping, emotional statements. "If you show AI text to editors, they will shake their heads because the content is empty. The clearest sign is that it contains many meaningless words. In journalism, we always say: 'Show, don't tell'," he explained.

Mahadevan also voiced concern that over-reliance on AI will make news reports resemble LinkedIn job postings, unsuitable for maintaining fairness and objectivity. "If journalists keep relying on it, local crime stories will start to look like police press releases, and climate change articles will echo corporate sustainability reports," he commented.

In March, The Atlantic magazine published "The Human Skill That Eludes AI" by journalist Jasmine Sun, examining machine writing capabilities. Sun noted that AI prose is full of errors, with chatbots constantly creating meaningless metaphors and repeating specific sentence structures mixed with a "fawning tone." This led her to question why LLMs, capable of memorizing a vast literary treasury, cannot write a decent essay.

After speaking with computer science scholars and individuals from AI writing startups, Sun realized that modern LLM models prioritize providing correct answers and strictly adhering to rules rather than writing well.

According to the article, LLMs undergo two training processes. In the pre-training stage, they absorb nearly all multimedia content on the Internet, compressing it into patterns. At this point, quantity outweighs quality because through large amounts of data, AI learns grammar rules, word associations, and predicts the next word in a sentence. Imperfections are further refined during the post-training stage, where models receive conversational samples to learn from and apply filters to avoid illegal requests.

Jasmine Sun believes that LLMs can only mimic human taste to a certain extent. She states that AI research is an experimental science that helps humans verify how an object functions. Art, however, does not follow rules and cannot be quantified. Initially, she thought AI's creativity was constrained by community standards and depended on the metrics of the companies that created it. Later, she concluded that AI models cannot write better because they lack lived experience and the ability to express emotions to offer unique perspectives.

On Literary Hub on 5/10, author Charlotte Donlon expressed similar views to Jasmine Sun. She argued that AI can absorb every book in a library and learn every writing style, but it cannot express emotional complexity or understand subtle yet weighty moments. According to Donlon, these belong to the soul, something AI cannot copy from humans. In the future, artificial intelligence will advance but will not develop the ability to feel pain or engage in moral introspection.

"If we want literary works that only humans can create, we must value the inner lives of writers as much as their work," the author stated.

Also in The Atlantic article, journalist Jasmine Sun interviewed James Yu, co-founder of Sudowrite, an AI platform that assists with fiction writing. While Yu expressed admiration for ChatGPT's progress over the years, he stated he would not want to read AI-written stories.

When asked why he thought AI could not write a truly good novel, he replied: "Most good debut works are autobiographical. Perhaps for a model to write a truly good book, it needs to have lived and died, to have truly experienced life."

Trinh Lam (according to The Atlantic, Reuters Institute, Literary Hub)