AI Content Detector
Paste any text to analyse whether it was written by AI or a human. Measures sentence uniformity, AI phrase patterns, passive voice, and more. Free, browser-based — nothing is sent to a server.
How AI Writing Differs from Human Writing
Large language models like GPT-4 produce text with distinctive statistical properties. The most reliable signal is burstiness — the variation in sentence lengths. Human writers mix very short sentences with long, complex ones. AI models tend to write sentences of surprisingly uniform length, producing text that flows smoothly but lacks the rhythm variation of authentic human prose.
AI models are also trained on a vast amount of formal, instructional text, which makes them over-represent certain phrase patterns: "it is worth noting", "plays a crucial role", "delve into", "moreover", and "in conclusion" appear far more frequently in AI output than in natural human writing.
Limitations of Heuristic Detection
No heuristic detector is reliable enough to use as a final verdict on its own. AI models are rapidly improving, and newer outputs are harder to distinguish. Human writing styles vary enormously — academic and formal writing shares many properties with AI output. Use this tool as a starting point for investigation, not as a definitive judgement.
What the Signals Measure
Sentence Uniformity
Measures the coefficient of variation in sentence lengths. Low variation (uniform lengths) is the strongest AI signal in this tool.
AI Phrase Patterns
Counts phrases like "it is worth noting", "delve into", "plays a crucial role", and similar expressions that AI models overuse.
Average Sentence Length
AI models cluster around 18–25 words per sentence. Very short or very long average sentence lengths suggest human writing.
Passive Voice
AI uses passive constructions ("is considered", "was designed") more frequently than natural human prose in most contexts.
Transition Overuse
Heavy use of "furthermore", "moreover", "consequently", and "nevertheless" indicates AI-style formal connective writing.
Score Interpretation
0–37: likely human. 38–61: uncertain or mixed. 62–100: AI signals present. Always interpret in context of the text type.
Frequently Asked Questions
How does this AI detector work?
The tool uses heuristic text analysis — not a trained machine learning model. It measures five signals: sentence length uniformity (burstiness), AI phrase patterns, average sentence length, passive voice density, and transition word overuse. Each signal is weighted and combined into a score from 0 to 100.
How accurate is it?
This is a heuristic tool — it is directionally useful but not definitive. AI-generated text often scores high due to uniform sentence structure and common phrases. However, formal academic writing, legal text, and technical documentation written by humans may also score high. Use the score as one signal among many, not as a final verdict.
What does "sentence uniformity" mean?
Human writers naturally vary their sentence lengths — mixing short punchy sentences with longer explanatory ones. AI language models tend to produce sentences of more uniform length. The tool measures the coefficient of variation (CV) of sentence lengths: low variation is an AI signal.
Why might a human-written text score high?
Academic, legal, and technical writing often uses formal language, transitions, and passive voice similar to AI output. A policy document, scientific paper, or business report may score high even if entirely human-written. Context matters: compare the score to the text type.
Can I use this for student essays?
The tool can provide a supplementary signal, but should never be used as the sole basis for an academic integrity decision. False positives occur, particularly for non-native English speakers who write more formally. Always combine automated signals with teacher judgement and conversation with the student.
What is the minimum text length?
The tool requires at least 40 words and 3 sentences to produce a result. Shorter texts do not provide enough statistical signal for any of the measured heuristics to be meaningful.