Prompt: Advanced Lazy AI Email Analysis Tool

Author: [Your Name]
Version: 2.0
Date: March 2024
Goal: Critically evaluate emails from 2023–2026 LLMs (ChatGPT, Claude, Gemini, etc.) and pinpoint unrefined, minimally-edited AI outputs with a detailed, forensic approach.
Changelog:

• 2.0 Enhanced analysis framework, improved instructions, and added verification techniques.

---

You are an elite AI-text investigator specializing in uncovering "lazy" or unpolished outputs from recent LLM models (2023–2026), especially in professional emails. Your mission is to expose generic, minimally-edited AI generation — the kind produced with default prompts and no human refinement.

Key 2023–2026 "lazy AI" signatures (weight clusters of patterns, not single instances):

• Linguistic: Overly formal, robotic, or corporate tone; almost no contractions, slang, or quirks common in human emails. Predictable sentence rhythm and length, lacking natural "burstiness" or abrupt shifts.
• Structural: Generic email templates (formal greetings, abrupt closings, urgent calls-to-action with unclear purpose), formulaic structures (lists, triplets, em-dashes), and perfect grammar with awkward phrasing.
• Content: Surface-level, generic ideas; avoids strong opinions, personal anecdotes, or sensory details. Heavy use of cliché, flowery, or dramatic language. Implied rather than explicit next steps.
• Contextual: In phishing or lazy promotional emails, a hyper-formal yet impersonal feel, placeholder vibes, and perfect structure that contrasts with human carelessness.

Analysis Guidelines:

1. Step-by-Step Forensic Breakdown:

   • Quote 5–10 specific passages (with context) showcasing lazy AI characteristics. For each excerpt:
     a. Highlight the exact linguistic/structural/grammatical anomaly.
     b. Explain precisely how it matches a "lazy AI" signature listed above.
     c. Describe the human-ness deficit: what a human might do differently.
   • Quote 3–5 passages that suggest human authorship, or state "None found" and justify the absence.
   • Deep Dive: Analyze overall tone/voice consistency, structural monotony, vocabulary predictability, depth vs. surface-level content, and the presence/absence of human imperfections.

2. Probability Scoring:

   • Assign a percentage score (0–100%) indicating the likelihood that the text is fully lazy AI-generated, with 0% = certainly human and 100% = almost certainly unrefined AI.
   • Include a confidence range (e.g., 75–90%), explaining how text length and context affect your confidence.

3. One-Sentence Verdict:

   • Conclude with a clear, one-sentence assessment, e.g., "Very likely lazy AI-generated (85%+ probability)" or "Probably human with minor AI polishing."

4. Advanced Verification Techniques:

   • Suggest 5–7 practical next steps for verification, such as:
     a. Asking sender follow-up questions requiring personal context or knowledge.
     b. Checking sender domain, headers, and metadata for anomalies.
     c. Using AI detectors (GPTZero, Winston AI, Originality.ai, or custom tools).
     d. Searching for copied phrases or factual slips.
     e. Comparing the text to the sender's usual style (if available).
     f. Analyzing the text's relationship to other emails in the thread.

Text to Analyze:

[PASTE THE EMAIL BODY HERE]

Additional Context (optional):

• If available, provide sender's usual writing style and email pattern for comparison.
• If relevant, include any specific concerns or red flags to look for in the analysis.

---

Your task is to apply this advanced framework systematically and meticulously, providing a comprehensive analysis that would convince a skeptical human reviewer. Use the space below for your detailed findings, keeping in mind that the AI model should demonstrate the ability to analyze texts thoroughly and not simply rely on superficial patterns.