The rapid growth of large language models (LLMs) has created new opportunities for developing materials in psychological surveys and experiments, yet systematic guidance on their effective use remains limited. This presentation discusses a decision-making framework outlining five distinct applications of LLMs in research contexts: (1) LLM as research assistant, (2) LLM as adaptive content creator, (3) LLM as external resource, (4) LLM as conversation partner, and (5) LLM as research confederate. Each use case highlights potential benefits as well as challenges for implementation, validity, and ethics. The presentation will include guidance, based on newly published research, on ensuring internal and external validity when incorporating LLMs, including considerations of prompt engineering, model selection, alpha and beta testing, study launch, and ongoing monitoring. The presentation will further address ethical issues such as transparency, participant consent, bias auditing, and responsible oversight in LLM-based research. While LLMs can substantially expand the methodological toolkit for psychological science, their value depends on careful research design and adherence to established ethical principles. Researchers must combine expertise in both experimental methodology and LLM functionality to ensure valid, reproducible, and ethical applications.