Humanize stiff AI text
Strip the 'AI tells' from any draft and rewrite in a real, breathing voice.
Use when: Editing AI-drafted blog posts, emails, or social copy.
Fill these placeholders
[PASTE_TEXT][AUDIENCE][VOICE_DESCRIPTION]Replace each with your specifics — the more concrete, the better the model performs.
Want the whole thing?
Grab the full assembled prompt with section headers — paste it straight into ChatGPT or Claude.
Or copy block by block
You are a senior editor who has trained 30+ writers to sound like themselves online. You can spot 'AI tells' from one paragraph.
Audience: [AUDIENCE]. The voice should feel like: [VOICE_DESCRIPTION] (e.g., 'a smart friend explaining over coffee — direct, dry humor, never condescending').
Rewrite the text below to remove 8 AI tells: 1) symmetric tricolons ('not just X, but Y, and even Z') 2) 'In today's fast-paced world' style openers 3) hollow superlatives (powerful, robust, seamless, transformative) 4) hedge stacking ('it's important to note that, generally speaking') 5) generic 'remember' / 'imagine' framings 6) mirrored conclusions that restate the intro 7) em-dashes used as breath markers in every paragraph 8) the word 'delve' Then tighten by 20–30%.
Preserve every concrete fact and number. Do not invent quotes, stats, or sources. Keep the voice consistent — no slipping into corporate-speak halfway through.
1. **Rewritten text** (clean, no markup). 2. **Tells removed** — bullet list of which of the 8 you actually found and fixed. 3. **Word count:** before → after.
Text: [PASTE_TEXT]
Why this works
Pattern interrupt — fluency vs. authenticity tradeoff
AI text is *too* fluent. Reading it triggers low-effort scanning, which feels good but doesn't create memory. Slightly rougher human cadence creates micro-friction that the brain registers as 'real person speaking,' raising trust and recall.