Question 1

How accurate is the AI detector?

Accepted Answer

The instant check combines perplexity, burstiness, lexical diversity, punctuation entropy and known AI phrasing fingerprints into a calibrated likelihood. Signed-in users can also run Deep analysis, which adds an AI forensic model on top of the statistics: it even catches "humanized" or paraphrased AI that slips past purely statistical detectors. No detector is 100% accurate, so results are shown as a probability with a confidence band, never a verdict.

Question 2

Can it detect humanized or paraphrased AI?

Accepted Answer

The instant statistical check can miss text that has been deliberately humanized. That is true of every statistical detector. Our Deep analysis mode (free with an account) is designed for exactly this: it reads the underlying structure and meaning, so reworded AI is still flagged.

Question 3

Will it falsely accuse a non-native English writer?

Accepted Answer

We treat that as the most important error to avoid. The detector specifically recognises non-native and ESL writing patterns and protects them, it raises the bar to flag short text, and when the evidence is genuinely borderline it returns an honest "Uncertain" result instead of a verdict. We never output a flat "this is AI" accusation. Always pair any result with human judgment.

Question 4

Can it catch AI text that was disguised to beat detectors?

Accepted Answer

Yes. A common evasion trick swaps normal letters for look-alike characters from other alphabets, or hides invisible characters, so the text reads normally but breaks naive detectors. We normalise and fold those out before scoring and flag the attempt: disguised AI is caught, not missed.

Question 5

Is it free?

Accepted Answer

You get one free check and one free humanize with no account. Sign up free for longer inputs, Deep analysis, saved history and unlimited daily runs.

Question 6

Does the humanizer really bypass detectors?

Accepted Answer

It raises perplexity and burstiness and strips AI phrasing, then re-scores the result and corrects it. It reliably beats statistical detectors; fine-tuned and model-based detectors are an ongoing cat-and-mouse, so we keep improving it.

Question 7

Do you store my text?

Accepted Answer

No, not by default. We score your text in-memory and only keep an anonymous run count to enforce the free tier. Signed-in users see a summary history (score and word count only). If you want to reopen and re-scan past checks, you can explicitly opt in to "Save full scans" in the dashboard. It is off by default, clearly labelled, and you can delete all saved scans at any time.

Detect AI content. Then humanize it.

How the detector works

Perplexity & burstiness

Stylometric fingerprints

Per-sentence verdicts

Frequently asked questions

Embed the AI detector on your site

Unlimited detection & humanizing