TelltaleProof — AI vs Human Content Detector

Deep, explainable AI content analysis across 30+ signals. Prove your content is human.

Phase 1 Features

• 8 scored sections, 30+ individual factors
• Bullet-point explanations for every score
• Aggregate score with verdict and confidence level
• Clean dark UI, fully mobile-friendly
• Powered by Claude API

Phase 2 Roadmap

• True perplexity scoring via Hugging Face GPT-2
• Supabase: user accounts, saved analyses, history
• NextAuth authentication + admin role
• Stripe billing (Free / Pro $19 / Agency $49)
• Shareable analysis URLs
• PDF export
• GitHub → Vercel auto-deploy

Setup

1. Install dependencies

npm install

2. Add your Anthropic API key

cp .env.local.example .env.local
# Edit .env.local and add your key:
# ANTHROPIC_API_KEY=sk-ant-...

Get your API key at: https://console.anthropic.com

3. Run locally

npm run dev

Open http://localhost:3000

4. Deploy to Vercel

npx vercel
# Add ANTHROPIC_API_KEY as environment variable in Vercel dashboard

Project Structure

app/
  page.tsx          — Main analyzer UI
  layout.tsx        — Root layout + metadata (SEO)
  globals.css       — Design system, CSS variables
  api/
    analyze/
      route.ts      — Claude API endpoint

components/
  ResultsDisplay    — Aggregate score hero + section list
  SectionCard       — Collapsible section with description
  FactorRow         — Individual factor with score bar + explainers
  ScoreGauge        — SVG half-circle gauge

types/
  analysis.ts       — TypeScript interfaces

Scoring Scale

• 0–3.5 → Likely Human (green)
• 3.5–6.5 → Ambiguous / Mixed (yellow)
• 6.5–10 → Likely AI-Generated (red)

Score Weights

Section	Weight
Cognitive Fingerprinting	16%
Word Choice & Phrasing	15%
Voice & Perspective	14%
Content & Logic	13%
Structure & Flow	12%
Emotional Texture	12%
Pragmatics & Subtext	10%
Statistical Proxies	8%