🔬
How it works
Measures sentence length variance (burstiness), vocabulary reuse patterns, AI transition phrases, and human-writing markers like contractions and first-person voice.
🎯
Sentence highlighting
Each sentence is individually scored. Pink = AI-like uniformity detected. Green = human writing markers present. White = neutral, insufficient signal.
⚡
100% private
All analysis runs in your browser using JavaScript. Your text is never sent to any server, never stored, never logged anywhere.
⚠️
Know the limits
No browser-side detector is 100% accurate. AI text that has been heavily edited, paraphrased, or humanised may score lower. Always treat results as a guide, not a verdict.
Frequently asked questions
How accurate is this AI checker?
This tool analyses statistical patterns common to all large language models — sentence length uniformity, vocabulary distribution and known AI transition phrases. It is most accurate on unedited AI text of 150+ words. Results on shorter texts, or text that has been manually edited after AI generation, are less reliable. Never use any AI detector as a sole source of truth.
Is my text stored or sent to any server?
No. All analysis runs entirely in your browser using JavaScript. Your text is not transmitted anywhere — it never leaves your device. It disappears as soon as you close the tab.
What is "sentence uniformity" in AI detection?
AI models tend to generate sentences of similar, moderate length — typically 15–25 words — with consistent structure. Humans naturally write short bursts, long flowing sentences, fragments, and asides. High sentence uniformity (low variance in sentence length) is a strong statistical signal of AI authorship.
What are "AI phrase patterns"?
Certain phrases appear disproportionately in AI-generated content: "furthermore", "it is important to note", "in today's world", "plays a crucial role", "in conclusion", "it is worth noting", and similar filler transitions. These are flagged and weighted in the analysis.
Can it detect text edited after AI generation?
Partially. If edits are minor, the underlying AI patterns remain. Heavy rewriting, paraphrasing tools, or running text through an AI humaniser can reduce the score. In practice, a score of 50%+ on edited text is still worth investigating.
Why did a human-written text score high?
Some human writers naturally write in a formal, structured style — particularly academic, legal or technical writers — whose output can resemble AI patterns statistically. Conversely, highly edited AI text may score low. The score is a probability estimate, not a definitive determination.