AI Tell Detector (EN + RU)

The problem

Polished AI-generated drafts often feel robotic due to predictable rhetorical patterns, symmetrical structures, and hollow fillers. Human readers and automated detectors catch these structural tells, making your content feel impersonal or untrustworthy.

What it does

• Identifies specific linguistic patterns like rhetorical symmetry, filler phrases, and repetitive sentence rhythms.
• Flags "hallucinated" personal experience and vague claims that lack concrete data or names.
• Analyzes structural "arcs" to detect overly clean transitions that signal AI involvement.
• Provides a precise audit table mapping problematic phrases to specific fixes or requests for human-only context.
• Detects "self-reflective winks" and formulaic call-to-action patterns common in LLM outputs.

Why this beats prompting it yourself

General prompts like "make this sound human" result in more vague AI-speak. This skill uses a 24-point technical checklist to isolate structural artifacts that generic LLM instructions consistently miss, such as parallel numerical triplets and clean "caveat blocks."

Use cases

• Reviewing LinkedIn posts or newsletters to remove marketing-speak and AI artifacts.
• Auditing professional emails to ensure personal anecdotes feel authentic rather than generated.
• Refining technical documentation to break up monotonous structural rhythms.
• Pre-processing content before running it through external classification tools.

Known limitations

Does not guarantee 0% detection on external tools, as classifiers vary wildly. Requires the user to provide missing factual details or real-life experiences when flagged.