Artificial intelligence tools have transformed many lives—for better or for worse. Data centers, which are both water- and energy-intensive, reportedly deplete groundwater and other freshwater resources while consuming vast amounts of electricity to remain operational. Recently, OpenAI’s CEO stirred controversy by suggesting that memory modules belong in AI data centers, dismissing their use in personal devices as a “waste.” This stance has contributed to shortages of DRAM and NAND chips, driving up prices for most consumer electronics. Now, it appears that AI is also altering the way people write; a report indicates that authors are shifting their writing styles out of fear of being falsely accused of plagiarism.
Authors Intentionally Introduce Errors to Evade AI Detection
Citing various research reports and accounts from several authors, Bloomberg reports that the emergence of AI detection tools has prompted writers to modify their writing styles and the structure of their work. The report highlights the case of author Mia Ballard, whose life took a sharp downward turn after Hachette Book Group canceled the publication of her book. This decision reportedly followed an assessment by the AI detection software “Pangram,” which assigned her book, Shy Girl, a 78.3% “AI-written” score.
Ballard has denied these allegations. She reportedly claims that the person editing her manuscript used artificial intelligence to polish the text without her consent. Subsequently, following the controversy surrounding Shy Girl—which involved the use of AI and triggered racist remarks online—Ballard self-published a novella titled Polly through a mental health organization.
Interestingly, Pangram detected that the first chapter of Poly was written entirely by artificial intelligence, despite the author’s claim that she had penned it herself. Similarly, the first chapter of her 2024 novel, Sugar, was flagged by the same detection software as being 83% AI-generated. In another novel, American Teen Princess, the author, Ballard, reportedly “dialed back” the poetic prose to make the writing sound more human. In this instance, Pangram returned a score of 37%, classifying it as a hybrid—a collaboration between a human and AI.
This is just one of many examples of how AI detection software is influencing the way humans write. Bloomberg cites researchers from the Max Planck Institute for Human Development in Germany, who noted a sharp spike in the use of the word “delve” in human writing following the release of ChatGPT. Furthermore, some writers have begun to avoid using em-dashes, as AI tools are known to frequently employ them when structuring sentences.
Nigerian author Jerry Falade has reportedly labeled these as “AI identifiers” or “AI hallmarks,” advising writers to steer clear of them. According to Bloomberg, Falade lost a multi-million-dollar book publishing contract after being accused of using AI to write his work. The emergence of AI detection tools has arguably led to the development of a “hybrid” style of language, where writers strive to avoid both the direct use of AI and any linguistic patterns that might make their work sound like it was generated by a machine. Conversely, some writers choose to use AI to draft content, subsequently “humanizing” the text through further refinement.
However, it is not just authors who are affected by the controversy surrounding AI detection. Several media outlets have dismissed journalists accused of using AI to write articles. There have also been instances where editors at tech news platforms were compelled to add disclaimers to articles after multiple people alleged that they had been generated by AI. Although the use of AI for writing—particularly creative writing—remains discouraged, it is currently unclear how AI detection tools can be refined to prevent the misidentification of human-written content as AI-generated.