AI Text Detector

Analyze text for AI writing patterns using heuristic signals. Not a definitive detector — for reference only.

Disclaimer: This tool uses heuristic analysis only and cannot definitively determine if text was AI-generated. Results are indicative — use for reference purposes only. No tool can guarantee 100% accuracy in AI detection.

Paste text to analyze

Paste at least 3-4 sentences for meaningful results

0 words

About AI Text Detection

This AI text detector uses heuristic analysis to identify patterns commonly associated with AI-generated content from systems like ChatGPT, Claude, and Gemini. It analyzes writing style signals rather than probabilistic language model outputs.

How It Works

Sentence Length Consistency: AI models tend to produce uniformly structured sentences (15–25 words), while human writing varies more.
Burstiness Score: Human writers naturally mix short punchy sentences with longer ones. Low variance (low burstiness) is an AI indicator.
Vocabulary Diversity (TTR): The type-token ratio measures lexical richness. AI tends to fall in a predictable mid-range.
Transition Words: AI overuses connective phrases like "furthermore," "moreover," and "additionally."
Passive Voice: AI-generated content often relies more heavily on passive constructions.

Limitations

No heuristic detector is foolproof. Academic writing naturally uses many AI-like patterns. A low score doesn't guarantee human authorship. Use this as one signal among many, not a definitive verdict.

How It Works

Sentence Length Consistency: AI models tend to produce uniformly structured sentences (15–25 words), while human writing varies more.
Burstiness Score: Human writers naturally mix short punchy sentences with longer ones. Low variance (low burstiness) is an AI indicator.
Vocabulary Diversity (TTR): The type-token ratio measures lexical richness. AI tends to fall in a predictable mid-range.
Transition Words: AI overuses connective phrases like "furthermore," "moreover," and "additionally."
Passive Voice: AI-generated content often relies more heavily on passive constructions.

Limitations

No heuristic detector is foolproof. Academic writing naturally uses many AI-like patterns. A low score doesn't guarantee human authorship. Use this as one signal among many, not a definitive verdict.