From Grammarly to anti‑Grammarly: Why ‘perfect’ now looks suspicious
Flawless writing used to signal care, competence, and education. Now, in inboxes and chat threads, it increasingly signals something else: that an AI model probably wrote it. As apps like ChatGPT and Gemini normalize polished, structurally sound prose, readers have begun to treat spotless emails with suspicion. Content professionals complain that work they labored over is being mislabelled as AI-generated because it shares the same fluent patterns detectors look for. Those systems often scan for consistent grammar, highly structured sentences, and certain stylistic tics common in large language models—traits that many skilled humans also exhibit. In response, some people have started doing the unthinkable on purpose: adding clunky phrasing, dropped articles, and mild typos so their messages feel more human. Into this cultural whiplash steps a new class of AI humanizer tools that deliberately roughen up text so it no longer reads like it came straight from a model’s output window.
How Sinceerly’s email typo plugin manufactures ‘human’ flaws
Sinceerly, a Chrome plugin created by Harvard Business School student Ben Horwitz, leans into this demand for imperfection. Rather than correcting your writing, it rewrites it to look less like it passed through a grammar gauntlet. The extension offers three modes. “Subtle” tightens text by trimming filler and turning phrases into contractions, typically slipping in a typo early in the email so it feels less polished. “Human” pushes further, adding a more conversational tone along with additional small errors. Then there is the striking “CEO” mode, which strips messages down to lowercase minimalism and aggressive brevity, sometimes even appending a casual “sent from my iPhone” if you forgot a signature. The idea is to mimic the fast, clipped style associated with busy executives whose power lets them ignore traditional etiquette—while reassuring recipients that a real person, not a model, is behind the keyboard.
The anti Grammarly AI trend and the rise of AI writing detection
Sinceerly has been described as an anti‑Grammarly AI: instead of sanding down rough edges, it introduces them. Its existence highlights a broader feedback loop between AI writing tools and AI writing detection. Detectors look for linguistic regularities—hyper‑consistent grammar, balanced sentence structures, and overused patterns—that often emerge from model training on vast text corpora. One analysis, for example, noted the boom in a particular “It’s not just X — it’s Y” construction, now so overrepresented in synthetic‑sounding corporate communications that it has become an informal tell. As these patterns are flagged, writers try to dodge them, sometimes by deliberately inserting irregularities. That, in turn, inspires tools that automate the chaos. What began as a push for clarity and correctness with traditional grammar checkers has evolved into a cat‑and‑mouse game, where being slightly wrong on purpose can make your writing look more credibly human than being perfectly right.
Office email, academic work and the ethics of looking human
In everyday office email, AI humanizer tools promise less stiff, more relatable messages. A carefully drafted note can be passed through an email typo plugin to gain contractions, softer phrasing, and minor blemishes that suggest spontaneity. Yet these tweaks are not harmless in every context. In professional negotiations or HR discussions, a stray typo about dates, figures, or responsibilities can create confusion or even conflict. In academic and corporate settings, using tools to disguise AI-authored passages raises deeper ethical concerns. Imperfection layers can become a form of deception, designed specifically to fool AI writing detection and human reviewers alike. That blurs lines of authorship and accountability: who is responsible for errors or misrepresentations in text that was drafted by one AI and “humanized” by another? As institutions refine policies around AI writing, intentionally obfuscating origin may become as fraught as plagiarism itself.
When to soften AI polish—and when accuracy should win
Used thoughtfully, AI humanizer tools can be helpful. Softening highly polished AI drafts for internal chats, quick status updates, or friendly outreach can make you sound less like a canned corporate memo and more like yourself. Light adjustments—shortening sentences, adding contractions, and loosening tone—are unlikely to cause harm. But when stakes are higher, clarity and precision should take precedence over sounding human. Contracts, policy documents, formal reports, and academic submissions should not rely on artificial typos to pass as authentic. A better strategy is to treat AI drafts as starting points: revise them heavily, inject personal context, and keep your own quirks rather than outsourcing them. If a tool’s main selling point is its ability to fool AI writing detection, pause and consider what you are trying to hide. Authenticity is not about sprinkling in errors; it is about being honest about how the text was made.