Ragam Bahasa Indonesia dalam Prompt AI: Studi Komparatif Gaya Respons ChatGPT

This study examines the influence of standard (baku) and non-standard (informal) Indonesian language in prompt construction on the response style and structure generated by ChatGPT. The proliferation of generative AI in Indonesia presents a gap: most users interact with AI using everyday informal language, while the effect on AI response characteristics remains understudied. Using a comparative qualitative approach, two prompt variants formal standard language and informal non-standard language were tested on an identical object: an internet service package (HOME ODS, 100 Mbps, Rp150,000/month). Responses were evaluated across five dimensions: structure, completeness, analytical depth, register alignment, and practical utility. Findings show that formal-language prompts yield more hierarchically organized and elaborated responses, while informal prompts elicit concise, conversational responses marked by emoji and colloquial tone. Both prompt types produced substantively comparable information, indicating that language variety primarily shapes response style rather than content depth. These findings suggest ChatGPT adapts its register to match user input, a behavior consistent with statistical pattern prediction inherent to large language models (LLMs). Implications for Indonesian language education and AI literacy are discussed.