git:20260918.40e721d to git:20261013.a69e0bb

18 added, 7 removed. Audit A to A.

---

name: content-locale-humanize-hi

title: "Hindi Native-Fluency Calibration"

description: "Per-language calibration anchors for detecting AI-slop and translationese in Hindi (hi) target text; auto-loads alongside content-locale-humanize when the target language is Hindi."

license: Apache-2.0

- compatibility: "Extends content-locale-humanize's AI-Slop & Translationese dimension. Load both together — this file has no rubric of its own."

+ compatibility: "Calibration data only; no tools required. Network access helps for usage checks."

domains: content

rules:

- content(hindi)

- match(\bhi-IN\b)

- match(\bHindi\b)

- match(हिन्दी|हिंदी)

- semantic(check if this Hindi translation sounds native)

- semantic(eliminate translationese from this Hindi text)

---

## Overview

- Sourced calibration anchors for Hindi, feeding `content-locale-humanize`'s AI-Slop & Translationese dimension. This is a calibration aid, not the checklist — reason natively beyond it (see the core skill's "why structure, not word lists" section).

+ Calibration anchors and current norms for Hindi (hi-IN), feeding content-locale-humanize's dimensions. A calibration aid, not the checklist; reason natively beyond it.

- ## Instructions

+ ## Rules

- Sourcing confidence: moderate — one specific, well-reasoned, distinctive finding rather than a broad convergent word list.

+ ### AI tells

- Over-Sanskritization: AI-generated Hindi systematically reaches for formal Sanskrit-derived (tatsama) vocabulary even in casual or practical contexts, where a native speaker would naturally use colloquial or Urdu-derived (tadbhava) words instead. Formal register applied to an everyday topic is itself a tell — check whether the vocabulary register matches the domain, not just whether individual words are "correct."

+ Sourcing confidence: moderate; one specific, well-reasoned finding rather than a convergent word list.

- Missing Hinglish: authentic contemporary Hindi — especially digital, informal, or conversational registers — code-switches with English words, phrases, and sometimes full clauses as a completely natural feature. AI-generated Hindi tends to avoid this entirely, producing an artificially "pure" Hindi that reads as stiffer and more formal than how the target audience actually writes or speaks. Flag content that never code-switches when the domain/register would call for it (casual blog, social, UI copy) as a likely AI/translationese signal.

+ - Over-Sanskritization: AI-generated Hindi reaches for formal tatsama vocabulary in casual or practical contexts where native speakers use everyday or Urdu-origin words. Check that vocabulary register matches the domain, not just that words are correct.

+ - Missing Hinglish: contemporary digital, informal, and conversational Hindi code-switches with English words and phrases naturally. Text that never code-switches in a casual blog, social post, or UI string likely reads as AI or translationese.

+ ### Translationese from English

+ - "द्वारा" agent passives ("यह फ़ाइल उपयोगकर्ता द्वारा बनाई गई") where Hindi uses the active ("उपयोगकर्ता ने यह फ़ाइल बनाई"); "किया जाता है" everywhere.

+ - English clause order kept intact instead of Hindi verb-final structure.

+ ### Current norms

+ - । (purna viram) ends sentences; lakh/crore grouping (12,34,567); follow house style on nukta (फ़, ज़) and on Devanagari vs Roman script for English terms, and hold it.

+ - Spelling follows the Central Hindi Directorate's standard orthography (मानक हिंदी वर्तनी).

## References

- gptcleanuptools.com (Hindi AI Detector product page, self-reported linguistic rationale — treat as a plausible starting hypothesis, not corpus-verified).

+ Reviewed 2026-10-12. AI tells: gptcleanuptools.com (Hindi AI Detector product page, self-reported rationale; a plausible hypothesis, not corpus-verified). Norms: Central Hindi Directorate (केंद्रीय हिंदी निदेशालय).