Artificial intelligence in research and academic writing: an analytical review of application domains and limitations

The rapid proliferation of artificial intelligence (AI) technologies into the domain of scientific research and academic writing necessitates a systematic analysis of their capabilities and limitations. This analytical review aims to identify key domains of AI application in scientific work ‒ from ideation and research planning to manuscript preparation ‒ and to determine the associated risks arising from the technological flaws of these tools. Based on a analysis of relevant scientific publications (2021–2025), seven core application domains were systematized: 1) ideation and planning; 2) literature search and review; 3) data analysis and statistical processing; 4) data visualization; 5) text drafting, editing, and reviewing; 6) text translation; and 7) speech and handwriting recognition and transcription. For each domain, functional capabilities are detailed, and characteristic limitations and risks are identified, including AI «hallucinations», the «black box» problem, the risk of unintentional plagiarism, disregard for cultural context, and data quality dependency, among others. The findings underscore the significant potential of AI to optimize research workflows, while emphasizing the critical need for researcher vigilance and continuous validation at all stages. Recommendations for the responsible use of AI are formulated, mandating the explicit disclosure of the artificial intelligence tools used in scientific publications, including their specifications and roles.