Intentionally bad writing and misspelling to prove that AI is not used

Hoang Nam, a senior marketing student in Ho Chi Minh City, finds that people are now quite sensitive to content that “looks perfect”. “A neat presentation file, a very smooth opening paragraph, a report with coherent steps are immediately asked ‘AI?’. It’s true AI, so I usually leave a few small errors to make it more natural,” Nam said.

Meanwhile, WSJ narrates the story of Sarah Suzuki Harvard, a professional writer in America, who does not tend to write plainly, but is changing. “I will use spoken language, like ‘hey, it’s true’, or add lots of exclamation points. Even though it’s uncomfortable, I still have to do that to sound human,” the 32-year-old content creator shared.

Harvard is afraid of being thought of as using AI to draft, after seeing that happen to many other authors and having to work hard to prove its “human” side. “It’s crazy. People are demanding proof of the unprovable,” she said.

Similarly, Garrett Marcy, a 28-year-old financial coordinator in Jacksonville, uses AI to support his work. “I’m not ‘innocent’, everyone uses this thing. If you use it to compose emails, you’re not impersonating any credit,” he said.

However, after giving the AI ​​commands, Marcy still edits to keep her own voice, such as removing long dashes (em dash), sometimes replacing them with self-written sentences even though she knows they are long, and even leaving spelling errors.

According to Andy O’Bryan, co-founder of an online community for entrepreneurs interested in AI, more and more people are “breaking” AI-generated documents by adding errors or writing a few awkward sentences. “Many times when reading a post on the content creation platform Substack or someone’s blog, you will see that in the middle of a perfect paragraph suddenly an obvious mistake appears,” he said.

Theo WSJas AI text floods the Internet, many people seek to discover which creators are using the technology to produce content. Therefore, authors who write all their works themselves and those who use chatbots for help have to learn a skill they have never bothered with before: writing like a human.

AI is a good writer because it is trained on a huge treasure trove of works, from classic literature to modern criticism. Models are also taught to write coherently and avoid unnecessary complexity. However, being too meticulous and using literary devices repeatedly can become a telltale sign of AI writing.

According to Ivan Jackson, founder of startup Writehuman, as models develop, it becomes difficult to distinguish text written by bots. His company’s software edits AI text to sound more human, even releasing monthly updates to keep up with trends. However, Jackson said, text written by humans is increasingly being flagged by AI detection tools, possibly because people unconsciously start copying the style of artificial intelligence.

The confusion between AI and human text causes a lot of controversy. In March, publisher Hachette recalled the book Shy girl (Wimpy Girl) after accusing part of the content of being written with AI, although author Mia Ballard affirmed that she did not directly use this technology. According to Ballard, an editor she hired used an AI tool to process the first edition of the novel.

In the past few years, many schools and teachers have also used AI detection software to find cheating students, but the effectiveness is not stable. According to ABC Newsin 2024, Australian Catholic University (ACU) recorded more than 5,000 cases of suspected academic violations related to the use of AI, but about a quarter of these were later cleared because Turnitin, a popular plagiarism checking software, encountered errors.

Theo Slateresearch by an expert from Hult International Business School said that Montclair State University (USA) once instructed lecturers to consider too good grammar as suspicious. The school advises: “Essays written by AI are often unusually accurate in grammar, word usage and editing.”

Despite the praise for artificial intelligence technology, some studies show that the phenomenon of “AI user discrimination” still persists. A 2025 Duke University survey shows that many people believe that their colleagues will think they are lazy and incompetent if they use AI, making them afraid to disclose this.

Since then, the trend of “bad writing” to prove not using AI was born, although not everyone supports it. Last week, user Akhil Salim expressed his protest on the content creation platform Medium: “I’ve noticed a trend in online text lately. It’s people intentionally writing with lots of spelling mistakes, basic grammatical errors, and weird fonts to show they didn’t create the text using AI.”

According to Salim, some authors even promote bad writing as a strength or proof of authenticity. “You cannot justify poor language skills in the name of ‘authenticity’. Please don’t do that,” he said.

Salim said he couldn’t even WhatsApp his friends in such a buggy style, let alone post publicly. “I find that trend offensive because I respect proper expression, whether in English or any other beautiful language on this planet.”