Method of Working with AI in Creating Scientific Articles: Algorithm, Criteria, and Practice

Abstract

This article presents a systematic methodology for using AI in scientific writing, based on the author's experience creating a series of ontological and informaciological works. The method includes a nine-step algorithm: from idea emergence to final publication, with quality control criteria (semantic consistency, terminological purity, absence of false claims, novelty, correct citation, hypothetical status marking). The dual role of AI is analyzed: technical assistant (literature search, systematization, formatting) and intellectual partner (proposing new terms, logic checking, finding analogs). The article honestly acknowledges limitations and risks: AI can generate hallucinations, does not understand meaning, and requires critical verification. Practical recommendations are offered for researchers. Based on analysis of current literature on human-AI collaboration frameworks (The Symbiotic Scholar, Dialogic triad), automated pipelines (PaperOrchestrator, Agent Laboratory, The AI Scientist), transparency works (Jeongbin Park on Human-in-the-Loop), and publishing policies (ICMJE 2026, ARTS, SIAM), an improved version 2.0 of the method is proposed. Enhancements include: dialogic querying, analysis of relationships with existing works, automated citation verification, ethical review, and a reflection step. The two-part structure allows readers to see both the empirically tested method and the theoretically grounded improved version. The key thesis: AI is a tool that enhances rather than replaces the author's thinking. The author remains responsible for meaning, position, and final text. The article concludes with an invitation to dialogue and collaboration in developing methodologies for AI-assisted scientific work. Keywords: methodology of science, artificial intelligence, LLM, scientific article, algorithm, quality criteria, verification, authorial method, AI-assistant, AI-partner, collaboration frameworks, publishing policies, transparency.