AI-Supported Remote Laboratories for Measurement Education: A Preliminary Evaluation of LLM for Oscilloscope-based Training

This work presents a preliminary evaluation of two available Large Language Models (LLMs), specifically Google Gemini 3 and ChatGPT-5.3 (i.e., as a chatbot), used as Artificial Intelligence (AI)-based assistants to support students during remote laboratory activities involving oscilloscope-based operation. The study focuses on the use of conversational AI to guide students in configuring a remotely accessible oscilloscope and performing manual measurements of signal amplitude and frequency using cursor-based techniques. The chatbots' performance is evaluated by comparing the generated procedural instructions with a reference laboratory procedure for manual instrument configuration and measurement execution. Multiple interaction sequences are analyzed to assess consistency, completeness, and operational usability of the generated guidance in realistic asynchronous learning conditions. Preliminary results show that the quality and correctness of the generated procedures significantly improve when prompts explicitly include interface-level interaction details and remote operation constraints. A comparison across independent test sessions reveals that while Gemini provides fairly consistent and similar responses, ChatGPT tends to generate different and increasingly detailed outputs when presented with the same task in new prompts. However, in both cases, when visual support is requested, the generated images are often generic and, in some cases, not directly relevant to the specific instrument interface, thereby reducing the effectiveness of the instructions during instrument configuration.