User Rating 0.0 ★★★★★

Total Usage 0 times

Category Text Analysis

Source Text 0 words

Linguistic Analysis

Sensitivity

Robotic Probability

--%

Based on linguistic fingerprints

Avg Entropy

Target: > 4.0 bits

Lexical Diversity

Unique Word Ratio

Sentence Entropy Map (Lower = More Robotic)

Analysis results will appear here...

Human (High Entropy)

Mixed

Likely AI (Low Entropy)

Is this tool helpful?

Your feedback helps us improve.

★ ★ ★ ★ ★

About

This tool utilizes stylometric analysis and information theory to detect linguistic patterns often associated with Large Language Models (LLMs). Unlike semantic detectors that look for factual inconsistencies, this engine analyzes the structure and predictability of text.

Artificial Intelligence models generate text based on probability maximization, often resulting in lower Shannon Entropy and uniform sentence structures. Human writing, conversely, exhibits "burstiness" - high variance in sentence length, complexity, and lexical diversity. By calculating the entropy density and lexical richness, we can flag text that lacks the chaotic nuances of human thought.

This detector operates entirely in your browser using local logic. It parses text into constituent sentences, calculates the entropy H for each, and maps the lexical diversity against a database of robotic fingerprints.

Formulas

The core mechanism relies on Shannon Entropy to measure the information density of a given text block X:

H (X) = - n \sum i=1 p (x i) log 2 p (x i)

Where p(x_i) is the probability frequency of character x_i. We also utilize the Type-Token Ratio (TTR) for lexical diversity:

TTR = V unique N total

The Robotic Score R is a weighted composite of entropy inverse and pattern matching:

R = α (1 - H norm) + β (Burstiness -1)

Reference Data

Metric	Human Baseline	AI / Synthetic Baseline	Significance
Shannon Entropy	High Variance (> 4.5 bits)	Low / Uniform (< 3.8 bits)	Measures the unpredictability of information content.
Lexical Diversity (TTR)	0.55 - 0.75	0.35 - 0.50	Ratio of unique words to total words. AI tends to reuse common tokens.
Sentence Variance	High (Burstiness)	Low (Flat)	Humans vary sentence length significantly; AI seeks an "average" length.
Perplexity Proxy	Spikes on nouns/verbs	Smooth distribution	Measures how "surprised" a model is by the text.
Connective Density	Context-dependent	Over-utilized	Excessive use of "Furthermore", "In conclusion", "However".
Syllabic Complexity	Varied	Standardized	AI often prefers simpler, high-frequency token structures.

Frequently Asked Questions

No linguistic tool can definitively confirm a specific model. Instead, this tool detects "signals" of synthetic generation: low entropy, repetitive sentence structures, and statistical predictability. High scores indicate text that "looks" robotic mathematically, but false positives are possible with formal academic writing.

Formal writing often follows strict structures, uses transition words heavily ('Therefore', 'Consequently'), and reduces ambiguity. These are also traits of LLMs. The "Burstiness" metric helps differentiate, as human formal writing still typically contains more structural variance than AI.

The threshold adjusts the sensitivity of the highlighter. A lower threshold (0.2) will only flag the most obviously robotic sentences. A higher threshold (0.8) will highlight any sentence that shows even slight uniformity. Use 0.5-0.6 for a balanced analysis.

No. All analysis (entropy calculation, tokenization, syllable counting) happens 100% in your browser's JavaScript engine. Your text never leaves your device.