ANALYSIS OF SCIENTIFIC INFORMATION SEARCH RESULTS USING ARTIFICIAL INTELLIGENCE PLATFORMS

Abstract

The article examines the likelihood of artificial intelligence (AI) platforms generating inaccurate, non-existent, or fictitious bibliographic descriptions of scientific publications. Attention is drawn to the possible limitations of using AI platforms when searching for sources of scientific information, based on the results of their verification. To achieve the best results in searching for scientific information, it is proposed to use a selective search method, which involves forming a search query according to clearly defined criteria of the researcher. To assess the relevance and reliability of scientific publications obtained from the search results of AI platforms, a comparative analysis of the top 20 search results was used. Scientific publications found using a typical search query were selected as the information base for the study. Basic criteria for evaluating the relevance of search results for information sources when using AI platforms were determined.The study contains an analysis of bibliographic data of scientific publications from the lists of references generated by the AI platforms ChatGPT5, Gemini, Grok, DeepSeek AI, and Cloude. A comprehensive assessment of scientific information sources was carried out using relevance assessment criteria for the selected group of keywords: “complex systems”, “digital systems”, “artificial intelligence” and “data analysis”. The existence of scientific publications was verified and confirmed using the scientometric databases of the National Library of Ukraine, Google Scholar, Research Gate, and higher education institution repositories. It has been proven that in order to collect reliable scientific information, it is necessary to use as many specialized electronic scientometric databases, electronic libraries, archives, etc. as possible.Practical recommendations have been developed for researchers on the use of AI platforms in scientific activities.