AI Detector
Free AI-Style Check — With the Signals Named, Not Hidden
Paste a text and get a read: a band, a 0-100 style-signal score, and the top 3 machine-style signals actually found in your writing. Then fix them by hand — or hand the same text to the humanizer with one tap.
Limits first
A diagnosis tool, not a verdict machine
No AI detector is definitive — including this one. The check reports observable style signals: template phrases, uniform sentence rhythm, fence-post transitions. It never claims to predict what GPTZero, Turnitin, or any external tool will say about the same text, because nothing can.
See the difference
What you paste, and what the check actually tells you — including the case where the answer is 'this reads human'.
How the free AI-style check works
Open @vustHumanBot with the button below (or send /detect inside the bot). The check runs in Telegram — no account, no email wall.
Paste the text you want checked (at least ~80 characters, so there is enough signal to measure). Send it.
Read the result: a band (reads mostly human / mixed signals / reads strongly AI-generated), a 0-100 style-signal score, and up to 3 concrete signals found in your text.
If the verdict bothers you, tap "Humanize this text" — the same text goes straight into the humanizer's standard rewrite, no re-pasting.
This tool handles
- A verdict band: reads mostly human / mixed signals / reads strongly AI-generated
- A 0-100 style-signal score you can track across your own revisions
- Up to 3 concrete signals, named and quoted from your text — not an opaque percentage
- English, Russian, and other languages — signals are reported in the language of your text
- Your first check free, no signup, straight in Telegram (/detect or the button above)
- One-tap hand-off of the same text into the VUST humanizer when you want the fix
Not in scope
- Predicting GPTZero, Turnitin, Originality.ai, or any external detector's score — different models, different thresholds; only they can speak for themselves
- Plagiarism / similarity checking — that's corpus matching, a different technology entirely
- Proof of authorship in either direction — no style check, including this one, is definitive evidence
- Very short texts — under ~80 characters there isn't enough signal to measure
Run the free check on a real paragraph first. The deep-dive below explains what each signal means, why detectors disagree with each other, and what to do about a false positive.
Frequently asked questions
Is this a real AI detector like GPTZero or Originality.ai?
It's an AI-style signal check: a language-model judge reads your text and reports how strongly its style resembles typical machine output, with the top signals named and quoted. It does not replicate, predict, or approximate the scores of GPTZero, Originality.ai, Turnitin, or any other external tool — those use their own models and thresholds, and only they can tell you their own verdict.
What do the three bands mean?
"Reads mostly human" means few or no strong machine-style signals were found. "Mixed signals" means some machine-typical patterns sit alongside clearly individual choices — the most common result for real-world text. "Reads strongly AI-generated" means multiple strong signals stack together, the density typical of unedited machine output. None of the bands is proof of authorship in either direction.
How accurate is it?
No accuracy percentage would be honest, so we don't publish one. All AI-text detection — commercial or free — is statistical style inference: scores swing between tools and re-runs, short texts read near-random, and formal or second-language writing gets false-flagged. The value here is the named signals, which are verifiable by you in your own text, not the number.
What does it cost, and what are the limits?
Your first check is free, right in Telegram, with no signup or email capture. From there, checking continues with your VUST plan. The paid product is really the fix, not the diagnosis: your first rewrite through @vustHumanBot is free too, then you keep going with a VUST plan.
Will it tell me what Turnitin will say about my essay?
No, and no third-party tool can. Turnitin's AI indicator runs on Turnitin's own model under your institution's settings; the only reliable test is that system itself. What this check gives you beforehand is the part those tools hide: which passages of your text carry machine-style signals and why.
What happens when I tap "Humanize this text"?
The exact text you checked goes to the humanizer's standard rewrite — sentence-length variance, transition diversity, removal of recurring AI phrasing — with your facts, names, numbers, and language preserved. It's the same paid humanizer flow you'd get by pasting the text yourself; the button just saves the paste.
Score came back too machine-flavored?
The humanizer rewrites the style and keeps your facts