The Russian developer Rucontext has updated its plagiarism detection system by integrating a new neural network designed for semantic text analysis. The tool is specifically engineered to identify paraphrased content by comparing the underlying meaning of sentences rather than relying on literal word matching, which often fails to detect rewritten text.
The need for this update stems from the widespread use of large language models for academic writing. According to a 2025 survey by the 'I am a Professional' association, 85% of Russian university students use AI tools, with 43% utilizing them to generate essays and research papers. Traditional systems typically miss paraphrased content if a third of the original text is altered.
The proprietary multilingual model calculates the probability of a paraphrase by analyzing sentence structure, synonym usage, and syntax. It remains effective even when lexical similarity drops to 50–60%. The system distinguishes between coincidental similarities and actual plagiarism, ensuring that different facts expressed in similar phrasing are not flagged incorrectly.
The platform provides transparent reporting, displaying the original source alongside the checked text with highlighted matches, source links, and page references. Maxim Dymkov, CEO of Rucontext, noted that the system is calibrated to favor the author in ambiguous cases to avoid false accusations. The update is currently available to all users under existing subscription plans.
