AI Content Detector

The Content Detector reads a passage and returns one of three labels: likely human, mixed, or likely generated. It also returns a single sentence of explanation. That is an estimate from a language model. It is not a laboratory test, not a watermark check, and not proof that a person did or did not write the words. The form states this before you submit, and the result repeats it. Do not use the label to punish a student, to reject a job applicant, or to accuse a writer in public.

Models recognize some habits of other models: even rhythm, tidy lists, a reluctance to be specific. People write that way too, especially when they are writing in a second language or following a template. A human paragraph that was edited by a model can look mixed. A generated paragraph that was rewritten by a person can look human. The label is a hint about the surface of the text, not a history of who touched it.

Paste the passage itself. A description of the passage, such as this looks like AI, is not useful. A few sentences is enough for a rough estimate. A single slogan is not. The page asks for a little more than a few words so the model has something to judge. Very long text is cut at the same limit as the other tools.

You can ask for the estimate's note in English or in Arabic. The label names stay in the simple English phrases above so they can be compared from one run to the next. The note follows the language you chose. Right-to-left layout applies when the note is Arabic.

There is no score out of one hundred on purpose. A precise-looking percentage would imply a measurement this page does not have. Mixed means the model would not pick a side. Likely generated means the surface looks automatic. Likely human means it does not. All three can be wrong on the next paragraph of the same document.

The detector uses the same text provider as the writer and the translator. If that provider is busy or not configured, you get the busy or the not-connected message and no label. The page will not guess a label locally. A guess without the model would be a different product, and it would still not be definitive.

Abuse rules apply to the passage you paste. The check is there to stop a few clearly harmful requests. It is not a judgment of whether the passage is true. A false product claim can be human-written. A careful true paragraph can be generated. Truth and authorship are separate questions, and this page only touches the second, weakly.

If you are an editor on AIMAPER, do not replace the editorial standard with this label. Full write-ups are published because a person edited them and they are long enough. A detector score is not part of that rule. Keep using the word count, the source check, and your own reading.