AI Content Detector

You paste text into the input field and click Analyze. The engine scores each sentence individually, measuring perplexity and burstiness against patterns typical of AI-generated writing. The result shows an overall AI probability percentage, a plain-language verdict, and a color-coded sentence list so you can see which specific sentences pushed the score up or down.

Detection accuracy is highest for English text, because the underlying models are trained predominantly on English writing patterns. Other languages are accepted and will produce results, but confidence scores may be less reliable. For mixed-language documents, the per-sentence breakdown still works, though English sentences within a foreign-language text will score more precisely.

The minimum is 50 characters, which is roughly one short sentence. Very short inputs do not give the model enough context for a reliable verdict. The maximum per scan is 10,000 characters. If your document is longer, split it into sections and scan each one separately.

Accuracy is strong for clearly AI-generated or clearly human-written text, but mixed content, heavily edited AI drafts, and very short passages sit in a gray zone where results should be treated as one signal among several. Use the per-sentence scores to understand where the uncertainty comes from rather than relying solely on the overall percentage.

Yes. After a scan completes, you can copy the full report to your clipboard with the Copy Report button. Logged-in users also have their last 20 scans saved automatically and accessible from the Previous Scans list at the bottom of the page.

Most scans complete in 10 to 20 seconds. Longer texts with many sentences take slightly more time because each sentence is analyzed individually rather than processed as a single block.

Significant human editing reduces the statistical fingerprints that the detector looks for. A lightly paraphrased AI draft often still scores high, but a thoroughly rewritten passage may score closer to human. This is expected behavior, not a flaw. The per-sentence breakdown helps identify which sentences retain the most AI-generation markers after editing.

AI chatbots are not calibrated detectors. When asked whether text is AI-generated, they pattern-match against style expectations rather than measuring actual statistical properties. Content Detector uses a dedicated analysis pipeline that scores perplexity and burstiness quantitatively, then maps those scores to per-sentence verdicts. The output is a structured, repeatable report, not a conversational guess.

For logged-in users, the scanned text and results are stored in your scan history so you can review them later. We do not share submitted text with third parties. If you prefer not to have scans saved, copy the report immediately after the scan completes and delete the entry from your history.

Submit continuous prose rather than bullet lists or tables, since the detector relies on sentence-level analysis. Paste the original text without stripping formatting that would merge sentences together. For long documents, scan paragraph by paragraph to see where the AI-generation score peaks. Single isolated sentences give less reliable results than paragraphs with three or more sentences.