AI Detector
Analyze text for signs of AI generation. This tool examines perplexity, burstiness, vocabulary richness, sentence structure, repetition patterns, and readability consistency to estimate the probability that text was written by AI.
Analyze Text
All analysis happens locally in your browser. Nothing is transmitted.
How to Use This AI Detector
Paste Your Text
Copy and paste the text you want to analyze into the input field above. You need at least 50 words for analysis, but 200+ words produce more accurate results.
Review the Score
The tool instantly analyzes six statistical signals and displays an overall AI probability score from 0% (likely human) to 100% (likely AI), along with a verdict.
Explore the Breakdown
Click on individual signals to see detailed explanations of each metric. Check which specific AI patterns were detected and how each signal contributed to the final score.
How AI Text Detection Works
Score = W1*Perplexity + W2*Burstiness + W3*Vocabulary + ...AI text detectors work by identifying statistical patterns that differ between human and AI-generated writing. While no single signal is conclusive, combining multiple independent signals produces a more reliable estimate.
Perplexity (25%)
Measures how predictable word sequences are. AI models generate text by selecting the most probable next word, resulting in lower perplexity. Human writers make more surprising word choices, creating higher perplexity.
Burstiness (20%)
Measures the variation in sentence length and complexity. Humans write in “bursts” -- mixing short, punchy sentences with long, complex ones. AI tends to maintain a more uniform sentence length throughout a passage.
Vocabulary Richness (20%)
Analyzes the Type-Token Ratio (unique words vs total words) and hapax legomena (words used only once). Human writing typically uses a wider and more varied vocabulary. AI text tends to cycle through a narrower set of common words.
Sentence Structure (15%)
Examines the diversity of sentence openings. Human writers naturally vary how they start sentences. AI-generated text often falls into patterns of starting sentences with the same words or structures repeatedly.
Repetition Analysis (10%)
Scans for repeated phrases (trigrams) and known AI-typical expressions like “it is important to note” or “in today's digital age.” These formulaic phrases appear far more frequently in AI text than human writing.
Readability Consistency (10%)
Measures how consistent the reading difficulty is across paragraphs. Humans naturally vary complexity -- some paragraphs are simple, others dense. AI tends to maintain a remarkably consistent readability level throughout.
Frequently Asked Questions
How does this AI detector work?
This tool analyzes six statistical properties of text: perplexity (word predictability), burstiness (sentence length variation), vocabulary richness (word diversity), sentence structure patterns, repetition of phrases, and readability consistency across paragraphs. AI-generated text tends to be more predictable, uniform, and repetitive than human writing. Each signal is scored and combined into an overall probability.
Is my text safe? Does it get sent to a server?
Your text never leaves your browser. All analysis runs entirely client-side using JavaScript. No text is transmitted to any server, stored in any database, or sent to any API. You can verify this by disconnecting from the internet after the page loads -- the tool will continue to work perfectly.
How accurate is this AI detector?
This tool uses statistical heuristics rather than machine learning models, which means it provides a probabilistic estimate rather than a definitive answer. It works best on English prose of 100+ words. Short texts, technical writing, code documentation, and non-English text may produce less reliable results. No AI detector is 100% accurate -- use this as one data point among many.
Why does it need at least 50 words?
Statistical analysis requires a sufficient sample size to be meaningful. With fewer than 50 words, metrics like sentence length variation, vocabulary diversity, and readability consistency cannot be reliably calculated. Longer texts (200+ words) produce more accurate results because the statistical patterns have more data points to analyze.
Can AI-generated text fool this detector?
Yes. AI text that has been manually edited, paraphrased, or written with specific prompts to mimic human style may score as human-written. Conversely, some human writing (especially academic or formal writing) may score as AI-generated because it shares similar statistical properties. This tool measures statistical patterns, not authorship.
What do perplexity and burstiness mean?
Perplexity measures how predictable or "surprising" the word choices are. Low perplexity means the text uses very common, expected word combinations -- typical of AI. Burstiness measures variation in sentence length and complexity throughout the text. Human writing tends to be "bursty" with a mix of short punchy sentences and longer complex ones, while AI writing is more uniform in sentence structure.
Privacy Notice
This AI detector runs entirely in your web browser. Your text is never transmitted to any server, stored in any database, or sent to any API. All six analysis signals are computed locally using JavaScript. Your content remains completely private. You can verify this by disconnecting from the internet and confirming the tool still works.