StrikePlagiarism.com vs. the New Generation of AI Content Creating Tools: First Test Results for DeepSeek and Grok – Times Higher Education
Practical insights from and for
academics and university staff
Everything you need for each step
of your study abroad journey
Sponsored by
As artificial intelligence continues to evolve, so too must the systems that protect academic integrity. StrikePlagiarism.com has recently benchmarked its AI content detection engine against two of the most advanced generative language models available today — DeepSeek, which became publicly available in the EU in January 2025, and Grok, introduced to EU users as a paid service in March 2025. The results once again confirm: we are ready for the future.
StrikePlagiarism.com
In our latest evaluation, StrikePlagiarism.com analyzed a set of academic-style texts generated using DeepSeek — a fast-growing, high-performance large language model known for its dense semantic style and narrative fluency.
Our detection engine identified the text as AI-generated with a 99% probability, supported by a 100% AI Content Indicator. These results demonstrate the system’s ability to recognize DeepSeek’s characteristic writing traits — including uniform syntax, inflated vocabulary layers, and unnatural structural rhythm — all of which deviate subtly from authentic academic writing.
Each flagged segment in the report is supported by a clear explanation of linguistic anomalies, risk assessment scales, and contextual examples. Institutions receive not only a verdict, but a transparent, interpretable diagnostic tool.
To assess conversational models, StrikePlagiarism.com also tested content created by Grok, developed by xAI. Unlike typical essay-writing models, Grok is optimized for dialogue and informal interaction — presenting a more complex detection challenge.
Nonetheless, our system confidently flagged the sample with a 99% probability of AI generation, revealing Grok’s synthetic signature: compact semantic units, reiterative phrasing, and consistent coherence. The AI Content Indicator confirmed elevated automation risk, and highlighted passages were annotated with structural and linguistic explanations.
These results confirm StrikePlagiarism.com’s capacity to accurately detect not only conventional academic-style AI writing but also content from newer, less predictable language models.
With verified detection rates of up to 99% for newly released models, StrikePlagiarism.com continues to set the standard for AI-generated content analysis. The system processes over 100 languages, supports multi-format submissions, and is trusted by ministries, universities, and examination authorities across the world.
We don’t just detect plagiarism — we define the frontier of ethical academic verification in the age of artificial intelligence.
Learn more: www.strikeplagiarism.com
Register for free
and unlock a host of features on the THE site
