As a professor of social work, I recently found myself in an embarrassing predicament. One of my undergraduate students submitted an assignment that struck me as suspiciously polished. The prose featured advanced vocabulary, sophisticated sentence structures, and a liberal use of em dashes to set off ideas. Convinced this must be the work of a artifical intelligence (AI), I confronted her about the possibility of AI assistance. She responded not with defensiveness, but with evidence. She brought in two handwritten journals from her high school years, each about one hundred pages long. Flipping through them, I saw page after page of her cursive script, brimming with the same advanced language and em dashes that had raised my alarm. (Yes, this essay uses em dashes. No, that does not mean it was written by a robot. This site has used them for years and sees no reason to stop now.)

It was a humbling moment. My assumption had been wrong, rooted in the pervasive paranoia that now shadows eloquent writing in academia.

Just five years ago, a well-crafted essay would have been cause for celebration. Professors like me would praise students for their command of language, viewing it as a sign of intellectual maturity and hard work. Today, the landscape has shifted dramatically. An articulate statement often triggers suspicion of AI usage. This reversal reflects the rapid integration of AI tools into everyday writing, turning what was once a virtue into a potential vice. In classrooms and scholarly circles, the default response to elegance is no longer admiration, but rather scrutiny.

This suspicion has led to curious overcorrections among writers. Some authors now deliberately insert minor errors into their work to signal human authorship. It is akin to a gifted student intentionally missing a question on the SAT to avoid accusations of cheating for a perfect score. I have heard colleagues admit to this tactic in emails and drafts, adding a misplaced comma or awkward phrase as a subtle badge of authenticity. As someone who once favored em dashes for their rhythmic emphasis, I now try to avoid them altogether. They have become the biggest casualties in this AI-era witch hunt, symbols of machine-like precision rather than stylistic flair.

The issue extends far beyond academic essays. On Apple devices, users can compose or revise emails with a single tap on Apple Intelligence in the top-right corner. Microsoft Outlook offers similar ease through CoPilot in the top middle, allowing instant rewrites. Tools like CoPilot and Grammarly operate in the background of any Microsoft document, providing real-time suggestions for revisions and sentence completions. These features do not just edit; they also offer pre-packaged thoughts, potentially eroding our ability to reason independently. What begins as convenience risks atrophying original thought, as writers lean on algorithms for inspiration rather than grappling with ideas themselves.

Ironically, the most reliable method to detect AI involvement remains AI itself. Educators and publishers turn to detection software, only to find these tools are far from infallible. This reliance creates a painful paradox, where machines police machines in a hall of mirrors.

Worse, AI detection has become embedded in a monetized industry. Companies profit from tools that flag content as AI-generated and are incentivized to over-identify in order to demonstrate value. This creates a self-reinforcing cycle: false positives justify continued sales and traffic, even as they erode trust in human authorship. In academia—where claims are meant to be evaluated through evidence, argument, and peer judgment—relying on proprietary, profit-driven detection tools to police scholarship undermines those very standards. It privileges those who can afford or navigate these systems, risking silencing capable writers whose work appears “too polished.”

Peer-reviewed studies highlight these flaws. For instance, researchers evaluated several AI detection tools and found high rates of false positives on human-written text. These tools demonstrated low specificity and struggled with advanced AI models, suggesting that significant refinements are needed before they can be trusted. Such biases could unfairly disadvantage international contributors, compromising fairness in publishing.

These findings underscore a broader crisis in academic evaluation. Detection methods exhibit inconsistencies that erode their validity. High false-positive rates flag legitimate work, stifling discourse. In my field of social work, where nuanced expression is key to exploring human experiences, this suspicion discourages eloquence. Students may dumb down their writing to evade scrutiny, prioritizing safety over sophistication.

The villainization of well-written statements also affects faculty. I now second-guess my own drafts, wondering if a turn of phrase might invite doubt. This self-censorship extends to peer review, where reviewers might dismiss submissions based on stylistic hunches rather than content. The result is a chilling effect on intellectual exchange, contrary to the university’s mission of fostering debate.

In reflecting on my student’s journals, I see a reminder of writing’s human essence. Those handwritten pages, with their imperfections and flourishes, embodied thought in motion. As AI reshapes our words, we must guard against letting suspicion “cook” the written word entirely. Instead, let us nurture a campus climate that values the mind behind the message, not the machine that might mimic it.