Editorial review
Inspect articles, drafts and publishing copy when a passage feels unusually uniform, repetitive or formulaic.
LEXICAL INTELLIGENCE / BROWSER STUDIO
Statistical AI Text Detection
LexyAI?
Paste the text. Read the pattern.
Free AI text detector · AI writing checker · ChatGPT text checker
LexyAI is a free AI text detector built around measurable writing behavior. It examines sentence rhythm, lexical diversity, word-frequency structure, repetition, punctuation and other signals entirely inside your browser, then combines them into an interpretable AI-pattern estimate. No account is required and the text-analysis engine does not need a text upload server.
10signals
0text uploads
1browser
∞free analyses
01Analysis Console
01
Extract
Words, sentences, punctuation, lexical variety and repetition patterns.
02
Measure
Ten independent statistical signals are normalized to a common scale.
03
Combine
A weighted heuristic produces an AI-pattern score plus a separate confidence measure.
02How It Works
Pattern over guesswork
What does an AI text detector actually measure?
AI-generated writing can contain statistical regularities in the way sentences are shaped, vocabulary is reused, phrases are connected and punctuation is distributed. None of these characteristics belongs exclusively to AI, but together they can describe the statistical texture of a passage.
LexyAI analyzes ten such dimensions. Sentence burstiness measures how much sentence length changes through a passage. Vocabulary metrics examine local lexical diversity and the rate at which new words appear. Word-frequency analysis looks at the shape of the frequency distribution. Repetition metrics inspect recurring adjacent phrases, while compression analysis provides a lightweight proxy for character-level regularity.
The result is intentionally presented as a pattern estimate. A high score means that the measured features collectively resemble the detector's AI-associated reference profile more strongly; it does not establish who wrote the text. A low score is not proof of human authorship either.
03Use Cases
Built for inspection
A practical AI writing checker for everyday review.
Content workflows
Use an AI-pattern score as one extra review signal across marketing copy, documentation, blog drafts and machine-assisted text.
Education & research
Explore measurable writing characteristics such as burstiness, lexical diversity, phrase repetition and vocabulary growth.
Private inspection
Analyze unpublished or sensitive drafts locally in the browser without sending the passage to an external detector API.
04Signal Map
Ten dimensions
A wider view of the writing pattern.
01Complexity Shiftwindowed entropy variance
02Rhythm Variationsentence-length coefficient of variation
03Word DistributionZipf fit / frequency structure
04Vocabulary RangeMATTR lexical diversity
05Phrase Varietyrepeated adjacent word pairs
06Text PredictabilityLZW compression proxy
07Vocabulary GrowthTTR decay smoothness
08Generic Phrasingconnector / stock phrase density
09Short Fragmentsmicro-sentence ratio
10Punctuation Rhythmpunctuation spacing variation
05Private By Design
Client-side analysis
Your text is analyzed where you entered it.
Your pasted text is tokenized and measured locally, and the worker returns only the numerical analysis data needed to render the result.
This design removes the need for a text-processing backend and makes LexyAI useful for users who prefer not to submit drafts to an external AI detector. Opening the website still creates ordinary requests for HTML, JavaScript, CSS and fonts, but the text entered into the analyzer is not required to travel through a LexyAI analysis API.
For the precise scope of the application's data handling and hosting requests, see the Privacy Policy.
06Interpretation
Read the result responsibly
AI detection is an inference problem.
A statistical text detector cannot observe an author's intent or reconstruct the history of a document. It sees the final text and measures properties of that text. Formal human writing can look highly regular; edited machine text can become less regular; short passages simply contain fewer observations.
That is why LexyAI separates AI pattern score from confidence. The score describes how strongly the measured features align with the engine's AI-associated profile. Confidence describes the internal agreement of those features together with the amount of usable text structure available for analysis.
Use the result as a review aid, not as an authorship certificate. For high-stakes decisions, combine automated analysis with human review and independent evidence. Read the full methodology for details on each signal.
AI Writing Detection Guide
Free AI text detector for ChatGPT and machine-generated writing
LexyAI is a free AI text detector for people who want to inspect the statistical characteristics of a passage without creating an account or uploading the passage to a remote detector service. It can be used as a ChatGPT checker, a general AI writing detector, a content review aid, or simply as a way to explore how measurable properties of writing change from one document to another.
When people search for an AI text detector, an AI writing checker or a ChatGPT detector, they often want two things: a useful signal and a clear explanation of what produced it. LexyAI takes a transparent route.
Many AI detectors focus on a single hidden classifier. LexyAI takes a more transparent route. It extracts multiple observable features from the text, including sentence-length variation, local entropy, lexical diversity, word-frequency structure, repeated bigrams, character compression, vocabulary growth, generic connector density, short sentence fragments and punctuation rhythm. Each feature contributes to the overall profile instead of being treated as a magic “AI word” detector.
How to use LexyAI
Paste a passage of at least 50 words into the analyzer and select Analyse. For the most useful reading, look at both the headline score and the signal breakdown below it. A result where several different structural signals point in the same direction is more informative than a result driven by one isolated feature. Longer passages generally provide more observations than short snippets.
Can LexyAI detect ChatGPT text?
LexyAI can identify writing patterns that are associated with its internal AI reference profile, which can include passages produced with systems such as ChatGPT. It cannot identify a specific model with certainty from text alone, and a passage that has been substantially rewritten may no longer preserve the patterns the engine measures. Likewise, human writing can sometimes produce the same statistical characteristics.
Why sentence rhythm is useful
One of the most intuitive features is sentence rhythm. Natural prose often changes pace: a short statement may be followed by a long explanation, an aside or a fragment. Highly uniform sentence lengths create a different statistical profile. LexyAI measures this variation using the coefficient of variation of sentence lengths, while keeping the metric as one component of a larger signal set.
Why vocabulary diversity matters
Vocabulary diversity describes how broadly a text uses words over local windows. A technical document may deliberately repeat terminology, so repetition alone is not evidence of AI. LexyAI therefore looks at moving-window lexical diversity and cumulative vocabulary growth to capture how word choice evolves through the passage rather than relying only on a raw unique-word count.
Why phrase repetition matters
Generated writing can contain recurring transitions and adjacent word combinations. The detector measures repeated bigrams and also checks the density of common connective phrases. This can highlight formulaic phrasing, but common English expressions are not inherently machine-written, so the features are interpreted collectively.
Why this AI text detector runs locally
Privacy is part of LexyAI's architecture. The analysis engine is loaded and processes the supplied text in the browser. The application does not need a text-analysis backend to compute the score. That makes it possible to inspect drafts without sending the actual passage to a LexyAI processing endpoint. Standard hosting and third-party asset requests can still occur when the page loads; those are covered by the Privacy Policy.
AI detector accuracy and limitations
No AI detector should be described as perfectly accurate. Detection depends on the text available, its length, its genre, its language characteristics and whether the original writing has been edited or paraphrased. LexyAI's percentage is therefore a model-driven heuristic estimate rather than a measured probability of authorship. The confidence value is designed to communicate internal signal agreement and available evidence, not to turn the result into certainty.
What a responsible AI text checker should tell you
A useful detector should make its limits visible. LexyAI is designed around that principle: show the score, show the signal profile, explain what is being measured, and remind users that automated text analysis is not proof. The tool is most valuable when it helps a reviewer ask better questions about a passage instead of replacing human judgment.
Try the free AI text detector or read the full LexyAI methodology to see exactly how the ten signals work.