toololis
🤖

AI Text Detector

Detect AI-generated text with rule-based heuristics

0 words

Free AI text detector — rule-based, private, unlimited

Toololis AI Text Detector analyzes whether text was likely written by ChatGPT, Claude, Gemini, or other AI models. We use the same statistical heuristics that power paid services like GPTZero and Copyleaks — but free, unlimited, and entirely in your browser.

How to use this tool

  1. 1

    Paste the text

    Drop 200+ words for reliable analysis. Shorter samples give less accurate results — that's a fundamental limit of heuristic detection.

  2. 2

    Wait for analysis

    The analyzer runs 5 heuristics in parallel — all in your browser, in milliseconds.

  3. 3

    Read the breakdown

    Overall AI-likelihood score plus individual metric breakdowns. Review which signals triggered.

The 5 heuristics we measure

1. Perplexity proxy
Measures word predictability. AI tends to use common, predictable words more often than humans. We compute the ratio of rare words (not in top 500) to total words.
2. Burstiness
Measures sentence-length variance. Humans mix short punchy sentences with long flowing ones. AI produces more uniform lengths. Higher variance = more human-like.
3. Vocabulary diversity (TTR)
Type-Token Ratio = unique words / total words. AI has moderate TTR; humans vary widely. Too high or too low both suggest machine-generation.
4. Repetition score
Counts how often 3-word sequences repeat. AI tends to reuse phrases. Humans vary their phrasing.
5. AI-tell phrases
Common AI linguistic patterns: "furthermore", "it is important to note", "in conclusion", "delve into", "moreover", "a testament to", "navigate the complexities".

⚠️ Important: No AI detector is 100% accurate

False positives are common. Formal writing (academic, legal, technical) often flags as AI because humans also use structured prose. Non-native English speakers are disproportionately flagged because their prose tends toward common words. Short texts (under 200 words) have high error rates regardless of detector.

Never use AI detection as proof of misconduct. Use it as a conversation starter, never a verdict.

What this tool is good for

  • Quickly sanity-checking suspected AI-written content
  • Self-reviewing your own writing for AI-like patterns if you used AI-assistance
  • Understanding the heuristic signals AI detectors use
  • Comparing with paid tools before paying for subscriptions

What this tool is NOT good for

  • Accusing students of academic misconduct
  • Rejecting job applications / submissions
  • Any high-stakes decision
  • Detecting paraphrased / hybrid human-AI content

Frequently Asked Questions

Is this actually using AI to detect AI?
No. We use rule-based statistical heuristics — the same approach GPTZero, Originality.ai, and most commercial detectors use. No AI API, no server calls, no cost.
How accurate is it?
Accuracy varies. On long (500+ word) clearly-AI or clearly-human samples: 75-90% agreement with paid tools. On short or mixed samples: 50-70%. No AI detector is reliable enough for academic or legal use — treat all results as probabilistic signals, not proof.
What heuristics do you use?
Five signals: (1) Perplexity proxy — ratio of common vs. rare words. (2) Burstiness — variance in sentence lengths (humans vary more). (3) Repetition — repeated n-grams. (4) AI-tell phrases — "furthermore", "it is important to note", "in conclusion", etc. (5) Vocabulary diversity — Type-Token Ratio.
Can AI text fool this detector?
Yes, trivially. Paraphrasing, mixing with human edits, or prompting the AI to vary sentence structure all reduce detection. This is why no detector is "reliable" — it's cat-and-mouse.
Is my text sent to a server?
No. 100% client-side. Your text stays in your browser. Run any content through without privacy concerns.
Should I use this to accuse someone of using AI?
Absolutely not. False positives happen constantly — especially with formal, technical, or non-native-English writing. Use this as a prompt for discussion, never as evidence.
What's the difference vs. paid AI detectors?
Paid tools (Copyleaks, GPTZero Pro, Originality.ai) use similar heuristics but tune them with millions of samples. Our version uses the same PUBLISHED techniques but without their proprietary training data. Free vs. $15-29/month.

You might also like

{ }

JSON Formatter

Format, validate & minify JSON in your browser

.*

Regex Tester

Test regular expressions in real-time

⇄

Base64 Encoder / Decoder

Encode or decode Base64 instantly

🔒
100% Privacy. This tool runs entirely in your browser. Your data is never uploaded to any server.