ChatGPT Detector
Wondering whether something was written by ChatGPT? Paste it here to see an estimate, highlighted passages and the patterns behind the result.
100% Free • No Login Required • No Registration
0 words · 0 / 15,000 characters
Common ChatGPT writing habits
Opening with a restatement of the question, “delve,” “tapestry,” “it's important to note,” tidy lists of three, hedged balanced views and a concluding summary are frequent tells — though newer models vary their style more.
Limits of ChatGPT detection
Prompting for a casual style, manual editing and paraphrasing all reduce detectability. No tool can prove ChatGPT authorship.
Before and after editing
Raw ChatGPT output usually carries the clearest signals: even sentence length, polished transitions, balanced arguments and a tidy summary at the end. After a person edits the text — adding examples, cutting filler, changing structure — the signal weakens, and results move toward “mixed.” That is expected and is one reason no detector can be certain.
How to check a suspected ChatGPT text
Paste at least 80 words, ideally 300 or more. Read the highlighted sentences and their reasons rather than only the percentage. Then look for evidence outside the text: does it match the writer's earlier work, did it appear suddenly without drafts, does it cite sources that actually exist?
ChatGPT and other chatbots
Claude, Gemini, Copilot and other assistants share many patterns with ChatGPT because they are built on similar technology. The detector looks for those shared patterns and does not identify which specific product produced a text.
Understanding the limits
No AI detector is 100% accurate. Results are probabilistic estimates and should never be used as the only evidence that someone used AI. Treat them as one signal alongside drafts, revision history, and a conversation with the writer.
Frequently asked questions
Does it detect GPT-4 and newer models?
It looks for patterns shared by modern models rather than one specific version, so results apply broadly but with uncertainty.
Can it detect ChatGPT text that was translated?
Translation changes wording and often weakens the signal. Results for translated text should be read with extra caution.
Can it prove someone used ChatGPT?
No. It provides an estimate with explanations. Proof requires other evidence, such as drafts or the writer's own account.
Is it free?
Yes. No sign-up and no paid tier.