ig-humanizer

1# Instagram Humanizer V3
2 
3Rewrites any caption or carousel slide text to remove the AI tells that human
4readers notice, and audits a finished caption against the 2026 Instagram
5checklist. Based on Wikipedia's "Signs of AI writing" taxonomy, the 2025-2026
6stylometry literature, our own Instagram caption corpus (n=284, prevalence
7only), and Instagram-specific patterns (the first-125 fold, the
8lowercase-casual caption register, sized hashtags, sends-and-saves structure).
9**V3 (2026-09):** recalibrated on 2026 evidence. Vocabulary is scored by
10density, em dashes are capped instead of banned (and the cap is generous here:
1129% of human captions use one), forced rhythm is now a tell instead of a fix,
12and there is an over-correction guard.
13 
14**What this skill does not do:** it does not make text "pass" GPTZero,
15Pangram, Turnitin or Originality. Those are trained classifiers keyed on the
16instruction-tuning style signature; prompt-style "sound like a real person"
17rewrites are caught 92-95% of the time, and light mechanical rewriting raises
18detectability. On caption-length text (under 300 words) detector scores are
19noise. The real value is elsewhere: expert human readers cite vocabulary (53%)
20and sentence structure (36%) as what gives AI text away, and on Instagram a
21caption that reads as a brand account earns neither the save nor the send.
22This skill removes what those readers react to.
23 
24## What changed in V3
25 
26Evidence tier in brackets: [strong] = replicated across 2+ independent
27…