I'm Kenyan. I don't write like ChatGPT, ChatGPT writes like me

Accusations of “AI Writing” and the Curse of Being Polished

• Many describe being accused of using ChatGPT simply for writing clearly, formally, or at length—especially students, non‑native speakers, support staff, and professionals used to structured prose.
• Readers increasingly treat typos, grammatical quirks, and informal tone as proof of “realness”; polished language triggers suspicion. Some now deliberately insert mistakes or flatten their style.
• Commenters argue it’s rude and intellectually lazy to dismiss a message by yelling “AI” instead of engaging with its content.

Kenyan / Colonial English and LLM Training

• Several Kenyans say their schooling explicitly rewarded “big” vocabulary, proverbs, metaphors, and rigid essay structures, descended from British “Queen’s English” norms.
• That style functioned as a class and “civilisation” signal, not just exam technique.
• People note the irony that Kenyan (and other African) workers helped train OpenAI systems, and now Kenyans are penalized for sounding like the models they helped refine.
• Others push back that modern LLM voice is closer to US LinkedIn / content‑mill English than to classic colonial or academic prose.

What ChatGPT Actually Sounds Like

• Described patterns:
  • Overly “punched‑up” paragraphs, constant mini‑mic‑drops, clickbaity subheads.
  • Verbose, hyperbolic formulations (“not just X, but…”), corporate/marketing vibe, and “word salad” that uses many words to say little.
  • Technically decent grammar and rhythm, but often empty of real insight.
• Some see this as identical to business‑school and big‑tech review writing; others insist truly good prose (including the article) feels more grounded, purposeful, and information‑dense.

The Em Dash, Heuristics, and AI Detectors

• The em dash has become a meme “tell” for AI, even though:
  • Many humans used it heavily long before LLMs.
  • OSes often auto‑convert “--” into an em dash.
  • Style guides prescribe different spacing around dashes.
• Several argue single features (dashes, connectors like “furthermore”) are weak signals; more reliable cues are overall rhythm, fluff, and vacuousness.
• AI detectors frequently misclassify human text (including this essay), and people uncritically asking one chatbot to judge another’s output are widely ridiculed.

Cultural and Educational Fallout

• AI‑generated “slop” raises the cost of reading: everyone now runs personal, often faulty, heuristics just to decide what’s worth attention.
• Artists, writers, and even YouTubers report similar suspicions about AI voices or visuals.
• Some embrace LLMs as tools to mass‑produce required bland prose (academic papers, corporate comms), arguing English was already “slop” in those domains.
• Others worry about a “post‑truth” environment where genuine evidence and authentic voices are easily dismissed as synthetic.