AI-written text leaves patterns that human writing usually does not. Our check looks for them in two ways and combines the results into one verdict: green, amber or red.
- Language structure: we analyze statistical factors in your text - how much your sentence lengths vary, how your sentences begin, phrases and transitions AI overuses (“it is important to note,” “Furthermore,” “delve”), how even your paragraphs are, how much passive voice you use, how smoothly one sentence leads into the next, and patterns in word choice and punctuation.
- Writing style: a trained AI model reads your text passage by passage and judges how each one is written as a whole. Title pages, tables and reference lists are skipped, so only your own sentences are read.
Every statistical factor was tested against papers written before ChatGPT existed, and the ones that wrongly flagged human writing were dropped. Your report marks the exact sentences and passages behind the result, so you see why, not just a score. No AI detector is 100% accurate, so treat the result as a guide, not a judgement.
The writing-style model counts toward the verdict for English and German texts of 400 words or more. For shorter texts, and for French and Spanish, the verdict rests on language structure.