← The AI Hype Audit — all 181 verdicts
PARTLY “97% of people can tell you're using ChatGPT & Claude” — the studies say it's closer to a coin flip
The claim“Did you know 97% of people can tell you're using ChatGPT & Claude? The structure, the tone, the too-perfect polish give it away every time. Here are 6 prompts to fix it.”
Split this one down the middle. AI-flavored writing does have recognizable tells — uniform structure, over-polish, stock phrasing — and “humanize this” prompts genuinely help. To the poster's credit, he actually delivered: the 6 prompts sit in pinned comments, no gate, and they're reasonable practice. But the 97% is invented. Peer-reviewed work puts human detection of AI text near chance: a meta-analysis found 49.9% accuracy for GPT-3 vs human text — a coin flip — with recognition around 57% in controlled studies even with training, and GPT-4 output was judged human 54% of the time. Newer models are harder to spot, not easier. No published study or survey supports 97%, and the post cites none. The stat is the hook, and the hook is fabricated — which is a strange way to open a post about writing people can trust.
What holds up
- AI writing tells are real — structure, over-polish, and stock phrasing are documented patterns
- The 6 prompts were actually delivered in pinned comments, ungated — the payload exists and is reasonable
- The underlying advice (edit AI output toward a human voice) is worth taking
What doesn't
- The 97% figure is unsourced and contradicted by the literature — human detection runs 50–57%, roughly chance
- GPT-4 text was judged HUMAN 54% of the time in testing — the trend is toward harder detection, not easier
- A fabricated stat opening a post about authentic writing is the oversell auditing itself
The catch
The post fakes a statistic to sell you prompts for not sounding fake — the true version (“most readers can't consciously tell, but generic AI voice quietly costs trust anyway”) was available and stronger.
How to actually do it
- Take the prompts — they're free in the pinned comments and the humanizing practice is sound
- Use the honest motivation: the readers who CAN spot AI voice are disproportionately your sharpest prospects, and generic voice costs trust even when it isn't consciously detected
- Before repeating any “N% of people” stat from social, ask for the study — no citation, no repetition
Useful prompts hanging off a fabricated statistic — keep the payload, kill the number.
- Confidence
- High
- Posted by
- an AI-writing creator who delivers his prompt payloads in pinned comments
We test hype for free. We build the real thing for a living.
Thirty minutes, no pitch — and you'll leave with something useful either way.
Book a call with Todd or start with the free Business Checkup →The Verdict Weekly
Three verdicts every Friday. Free forever, unsubscribe anytime, no spam — that would be ironic.
© Schreier Group · schreiergroup.com · See a wild AI claim? Drop it here and we'll test it.