Copyleaks AI Detector & ChatGPT Plagiarism Checker - Apify Actor

Enterprise neural AI text detection & LLM authorship forensics Actor powered by the Copyleaks v10.0 classification model. Accurately analyzes text to verify human vs AI authorship across ChatGPT (GPT-4o, GPT-5, o1/o3), Claude 3.5 Sonnet, Google Gemini, DeepSeek-R1, Llama 3, and Quillbot with phrase-level character forensics, selectable sensitivity levels (1/3, 2/3, 3/3), and exact percentage scores.

---

🚀 Key Features

• Multi-Model LLM Detection: Detects text generated by OpenAI (GPT-3.5, GPT-4, GPT-4o, GPT-5, o1, o3), Anthropic (Claude 3.5 Sonnet, Claude Opus), Google (Gemini 1.5 Pro/Flash), DeepSeek (DeepSeek-R1, V3), Meta (Llama 3, Llama 3.1, Llama 3.3), and Quillbot.
• 3 Selectable Sensitivity Levels: Supports 1/3 (Low - Strict False Positive Prevention), 2/3 (Medium - Balanced / Standard Default), and 3/3 (High - Aggressive & Paraphrased Detection).
• Exact Probability Scoring: Returns precise ai_score (0–100%) and human_score (0–100%) alongside clear classification verdicts (AI_GENERATED, MOST_LIKELY_AI, or HUMAN_WRITTEN).
• Phrase-Level Character Forensics: Returns character offsets, snippet highlights, and AI anomaly intensity multipliers for detected synthetic patterns.
• Short-Text Auto-Adaptation: Seamlessly evaluates short snippets, single sentences, and long-form essays without minimum length restrictions.
• Flattened Dataset & Key-Value Output: Automatically saves the full JSON forensic payload to Key-Value Store (OUTPUT) and outputs flattened, structured rows to the default dataset.

---

⚡ Multi-Model AI Classification Coverage

This Actor inspects submissions against all major artificial intelligence models:

1. OpenAI Foundation Models

• Supported: ChatGPT, GPT-4o, GPT-4o-mini, GPT-4 Turbo, GPT-5, OpenAI o1, OpenAI o3, and legacy GPT-3.5.
• Output Forensics: Identifies characteristic GPT structural transitions, repetitive sentence pacing, and formulaic conclusions.

2. Anthropic Claude Models

• Supported: Claude 3.5 Sonnet, Claude 3.5 Haiku, Claude 3 Opus, and Claude 4.
• Output Forensics: Detects Claude-style nuanced hedges, high-vocabulary corporate phrasing, and analytical reasoning outlines.

3. DeepSeek Reasoning Models

• Supported: DeepSeek-R1, DeepSeek-V3, and DeepSeek Coder.
• Output Forensics: Flags algorithmic step-by-step reasoning structures and technical thesis breakdowns.

4. Google Gemini & Meta Llama

• Supported: Gemini 1.5 Pro, Gemini 1.5 Flash, Llama 3 (8B, 70B, 405B), Llama 3.1, and Llama 3.3.
• Output Forensics: Detects Gemini conversational phrasing and Llama token frequency distributions.

5. Paraphrasers & AI Spinners

• Supported: Quillbot, WordAI, Undetectable AI, and StealthGPT.
• Output Forensics: Uncovers heavy synonym replacements and modified synthetic syntax designed to evade naive filters.

---

🎯 What People Use It For

• Academic Integrity & Higher Education: Screen student admissions essays, research papers, literature reviews, and homework assignments for generative AI contributions.
• SEO Content Auditing & Publishing: Audit freelance copywriters, blog posts, affiliate reviews, and news articles to maintain domain authority and avoid Google search ranking penalties.
• Recruitment & Talent Screening: Verify candidate cover letters, take-home writing assignments, and technical interview responses.
• Legal & Due Diligence Review: Verify corporate communications, compliance filings, and contract summaries.
• Marketplace & Forum Moderation: Screen user-generated content, fake product reviews, and automated bot discussions.

---

📌 Decision Rule & Usage Recommendation

Check if success === true AND evaluate ai_score:

• ai_score >= 50: Classify as is_ai_generated: true with verdict AI_GENERATED or MOST_LIKELY_AI.
• ai_score < 50: Classify as authentic human authorship with verdict HUMAN_WRITTEN.
• Inspect patterns_detected: Iterate through the patterns_detected array to highlight exact character positions and phrases written by AI.

---

💡 What Problem Does This Solve?

• "Generic AI detectors give inaccurate false positives on authentic human writing."
• "Standard commercial AI detection APIs cost hundreds of dollars per month with heavy enterprise commitments."
• "Most detectors fail on short sentences or require at least 350+ words to run."
• "Content managers need exact phrase-level highlights, not just a vague overall percentage."

Copyleaks AI Detector delivers automated, high-precision AI content forensics in a single Actor run.

---

🔍 Queries This Actor Answers

• How do I check if an essay or article was written by ChatGPT or Claude programmatically in Python, Node.js, or Apify?
• What is an accurate API alternative to Turnitin AI, GPTZero, and Winston AI?
• How do I extract exact phrase character offsets and AI scores from text submissions?
• How do I detect Quillbot paraphrased AI text automatically?

---

📥 Input JSON Examples

Example 1: Standard AI Content Detection (Copyleaks v10, Sensitivity 2/3)

{
  "text": "In today's fast-paced digital era, the convergence of artificial intelligence and machine learning represents a transformative milestone in modern technology. Generative language models, such as GPT-4 and Claude, leverage deep transformer architectures to process vast datasets and generate human-like prose with remarkable coherence.",
  "engine": "copyleaks-v10",
  "sensitivity": "2/3"
}

Example 2: Authentic Human Writing (Sensitivity 1/3 Strict)

{
  "text": "I spent Sunday morning in my grandfather's workshop restoring an old oak rocking chair. We sanded down the weathered armrests by hand using medium-grit sandpaper before applying two coats of warm linseed oil. It was hard work, but seeing the natural wood grain come back to life was incredibly satisfying.",
  "engine": "copyleaks-v10",
  "sensitivity": "1/3"
}

---

📤 Dataset Output Format Examples

Output 1: AI Generated Content (ai_score=100%)

{
  "success": true,
  "engine": "copyleaks-v10",
  "sensitivity_level": "2/3",
  "text": "In today's fast-paced digital era, the convergence of artificial intelligence and machine learning represents a transformative milestone in modern technology. Generative language models, such as GPT-4 and Claude, leverage deep transformer architectures to process vast datasets and generate human-like prose with remarkable coherence.",
  "text_length": 342,
  "word_count": 42,
  "human_score": 0,
  "ai_score": 100,
  "is_ai_generated": true,
  "verdict": "AI_GENERATED",
  "confidence": "HIGH",
  "model_version": "v10.0",
  "scan_id": "scan_8fa732bc",
  "patterns_detected_count": 1,
  "patterns_detected": [
    {
      "index": 1,
      "snippet": "In today's fast-paced digital era, the convergence of artificial intelligence",
      "ai_intensity": 10.0
    }
  ],
  "scanned_at": "2026-09-01T20:30:00.000Z"
}

Output 2: Human Written Content (human_score=100%)

{
  "success": true,
  "engine": "copyleaks-v10",
  "sensitivity_level": "2/3",
  "text": "I spent Sunday morning in my grandfather's workshop restoring an old oak rocking chair. We sanded down the weathered armrests by hand using medium-grit sandpaper before applying two coats of warm linseed oil. It was hard work, but seeing the natural wood grain come back to life was incredibly satisfying.",
  "text_length": 308,
  "word_count": 52,
  "human_score": 100,
  "ai_score": 0,
  "is_ai_generated": false,
  "verdict": "HUMAN_WRITTEN",
  "confidence": "HIGH",
  "model_version": "v10.0",
  "scan_id": "scan_4c91a0ef",
  "patterns_detected_count": 0,
  "patterns_detected": [],
  "scanned_at": "2026-09-01T20:30:00.000Z"
}

---

⚠️ Failure Context & Error Handling

If the text parameter is omitted or empty:

{
  "success": false,
  "error": "Missing required parameter: 'text' cannot be empty."
}

---

🏷️ Keywords

copyleaks ai detector, chatgpt detector api, ai content detector, claude ai detector, gemini detector, deepseek detector, plagiarism checker, ai text forensics, quillbot detector, turnitin alternative, gptzero alternative, apify actor.