AI Watermark Detector (Text)
Heuristic signals for synthetic-looking text (entropy, repetition, distribution). Not a definitive detector.
Last reviewed by the Radiatus Cloud team
Output
Securing AI in production?
We build guardrails, governance & compliance for AI systems.
Look for signals of synthetic text
Some AI text carries statistical signals that hint at machine generation. This tool provides heuristic signals for synthetic-looking text, based on entropy, repetition and distribution, as indicators rather than a definitive verdict.
Why these signals are only heuristics
Machine-generated text can differ statistically from human writing in measurable ways, lower surprise between words, certain repetition patterns, distributions that skew toward the most probable choices, and these heuristics surface such signals. But they are genuinely just heuristics: capable models produce text that evades them, human writing sometimes triggers them, and editing changes everything. So the output is a set of indicators to weigh, never proof, and using it to accuse someone of using AI would be misusing it. The honest framing is a weak signal among many, and that limitation is worth stating clearly.
Indicators to weigh
The tool runs entirely in your browser, so nothing you paste, prompts, outputs or documents, is uploaded, which matters when the input is sensitive AI data or your own content.
Related tools
- AI Prompt Leakage Analyzer — Paste a system prompt and a hostile user input to see whether the prompt holds secrets and whether the input carries injection patterns. Local, instant.
- LLM Data Exposure Checker — Check if text contains data likely to be memorized or exposed by LLMs.
- AI Usage Policy Generator — Generate an acceptable use policy for AI tools in your company.
- Model Hallucination Estimator — Estimate risk of hallucinations based on task type and temperature.
Frequently Asked Questions
What signals does it look at?
Heuristics like entropy (the surprise between words), repetition patterns, and distributional skew that can differ between human and machine text.
Is it a reliable detector?
No. These are heuristics only. Capable models evade them and human writing can trigger them, so the output is indicators, not proof.
Can I use it to prove text is AI-written?
No. It should never be the basis for an accusation. It is a weak signal to weigh among others, not a verdict.
Why are these signals imperfect?
Because AI text quality keeps improving and human writing varies widely, so no statistical signal cleanly separates the two.
Is my text uploaded?
No. The analysis runs entirely in your browser.
Privacy & Security
Processed locally in your browser. No data is sent to any server.
How to Use
Paste text to compute simple statistical signals.
Disclaimer: This tool is provided "as is" without warranty of any kind. Results are for educational and utility purposes.
Related Tools
AI Prompt Leakage Analyzer
AI SecurityPaste a system prompt and a hostile user input to see whether the prompt holds secrets and whether the input carries injection patterns. Local, instant.
LLM Data Exposure Checker
AI SecurityCheck if text contains data likely to be memorized or exposed by LLMs.
AI Usage Policy Generator
AI SecurityGenerate an acceptable use policy for AI tools in your company.