A “Sanitary Perimeter” for Science

When Every Paragraph Is Under Suspicion

On January 12, 2026, the Russian EdTech service Domate announced a major update to its algorithm for detecting machine-generated text. The announcement comes at a moment when large language models (LLMs) are rapidly permeating education and research workflows.

At the core of the update is a systemic rethink of how AI-generated content is identified. The Domate team reports higher detection coverage, greater robustness across text lengths – from short excerpts to full dissertations – and targeted adaptation to academic writing. This distinction matters: scholarly texts differ fundamentally from student essays or news articles in structure, vocabulary, and argumentative logic.

The significance of this development extends far beyond a single product. As AI-assisted writing becomes widespread in term papers, theses, and even journal submissions, a new infrastructure for academic integrity is taking shape. Universities, publishers, and dissertation committees increasingly face the same question: how can human reasoning be distinguished from machine imitation? Domate positions its solution not simply as a “detector,” but as a trust-building tool that reduces institutional reputational risk while minimizing the likelihood of false accusations against students.

From a Check Button to System-Level Integration

Over the next one to two years, the sector is likely to move from fragmented tools toward deep integration of AI detection into educational and research ecosystems. Domate is betting on usability: demo access, bots, and APIs point to a strategy of broad adoption through familiar channels. A practical scenario is easy to imagine: an instructor uploads an assignment to an LMS and immediately sees not only a plagiarism score, but also annotated zones of potential AI generation, accompanied by explanations.

We see how quickly AI tools are being integrated into academic writing, creating new challenges for existing detectors. At one point, it seemed there would be no winners in this race. Today, however, we are ready to say that the solution presented here sets a new standard for AI detection in scientific and educational materials

Yury Chekhovich

Executive Director of Antiplagiat and founder of the Domate academic integrity service

Explainability is set to become the defining trend. The market has grown wary of black-box systems that output “37% AI” without context. The future belongs to solutions that show where, why, and with what degree of confidence machine authorship is suspected. This approach reduces disputes, protects good-faith authors, and makes review processes more transparent.

The segmentation between education and research is particularly notable. Scholarly writing is not just prose; it is formal logic, citations, and methodology. By focusing on the specifics of Russian-language academic discourse, Domate may occupy a niche where universal Western detectors often struggle.

Five Years of a Race – From Illusions to Realism

A look back shows how uneven this path has been. In April 2023, Turnitin, the global leader in plagiarism detection, launched AI detection and effectively legitimized the technology in global education. Yet by July of the same year, OpenAI shut down its own text classifier, citing insufficient accuracy. The episode underscored a hard lesson: AI detection is not magic, but a technically complex task with a high risk of error.

In Russia, a system-level milestone came with the rollout of AI detection in the Antiplagiat service in May 2023. Universities, from regional institutions to Moscow State University, began drafting internal rules governing the use of such tools and warning students about false positives. At the same time, the research community intensified work on specialized datasets – including the RU-AI project in 2024 – to improve reliability for the Russian language.

Those Who Integrate Best Win

Domate does not promise 100% accuracy – and that restraint is deliberate. In the “detector race,” success favors those who deliver stability, interpretability, and seamless integration rather than bold claims. The emphasis on Russian-language scholarly writing is especially valuable: Western models often produce inconsistent results on non-English corpora, creating an opening for local developers.

Export potential is limited but tangible – in countries with strong Russian-language academic communities, within diasporas, and among international publishers working with Russian-speaking authors. The key prerequisite is transparent methodology supported by open benchmarks.

Ultimately, tools like this are less about surveillance than about preserving trust in knowledge itself. In a world where AI writes more and more, the ability to distinguish human thought from its imitation becomes essential. Domate is taking another step toward ensuring that science remains a human endeavor, even in the age of artificial intelligence.