Cereby
Sign inGet started free

Free AI Text Detector

Check if text was written by AI. Paste up to 2,000 characters. We run the same detector that powers Cereby's paid app and return a 0 to 100 score, a verdict, and the signals that drove it.

0 / 2,000 characters · Your text is not stored

How to use this tool

  1. 1

    Paste the text you want to check (between 50 and 2,000 characters).

  2. 2

    Solve the quick verification challenge.

  3. 3

    Click 'Scan for AI'. The detector runs on our server and returns in a few seconds.

  4. 4

    Read the score, the verdict, and the signal breakdown.

Why this tool exists

Modern AI text classifiers look for statistical fingerprints, not specific words. They learn that AI output tends to be smoother, more predictable, and structurally uniform compared to human writing, which is rougher and more varied. The score reflects how strongly your text matches the AI side of that distribution.

A few things to keep in mind. First, scores are not absolutes. A polished, formal piece by a careful human can score high; a lightly edited AI draft can score low. Second, signals are explanations, not proof. When the detector flags 'high em dash density', it means that pattern is common in AI output, not that em dashes are inherently AI. Third, classifiers update. What scored 90 last quarter might score 60 today as the detector and humanizers evolve in parallel.

The most useful workflow is iterative: detect, rewrite the highest-signal sections, detect again, see what moved. The free tool gives you the detect side of that loop. The full Cereby app pairs it with a rule-based humanizer, per-paragraph scores, and detector feedback that tells the humanizer which signals to attack on the next pass.

When you'd use this

  • Sanity-checking a draft before turning it in
  • Spotting AI-written sections in a longer piece you're editing
  • Validating that your humanized output now reads as human
  • Quickly testing how a detector reacts to different rewrite styles

Frequently asked

What detector does this use?
It uses the same in-house Cereby detector that powers our paid app. No third-party API is called, so your text never leaves our infrastructure.
Why is the limit 2,000 characters?
Two thousand characters is the natural input window our detector handles in a single pass without truncation. Anything longer gets cut off, so we cap at the limit where the score reflects everything you pasted. The full Cereby app splits long documents into paragraphs and detects each one separately.
Why does very short text score 0?
Below 50 characters there isn't enough signal for the detector to make a confident call, so it returns a neutral 0 instead of guessing. Paste at least a couple of full sentences for a meaningful score.
Can a high score guarantee text is AI-written?
No detector can guarantee that. A high score means the text shares strong statistical patterns with AI output (predictable rhythm, em dashes, formal connectors, low burstiness), but careful human writing can score high too, and edited AI text can score low. Use it as a signal, not a verdict.
What do the signals mean?
Signals are specific patterns the detector and our analysis layer flag in the text: em dash density, formal connector frequency, sentence rhythm uniformity, vocabulary register, and contraction usage. They're the same signals our humanizer attacks when rewriting.
Is my text saved or used for training?
No. The text is sent to our detector, scored, and discarded. We don't store it, log it, or use it to train any model.
What's different in the paid Cereby app?
The paid app removes the 2,000-character cap, scans documents paragraph-by-paragraph with per-section scores, integrates the detector with the humanizer for an iterative rewrite loop, and tracks scores across drafts so you can see what's improving.

More free tools