Prompt Engineering as a New Literacy: The Linguistics of Talking to Large Language Models

Abstract

Since the release of large-scale generative language models, the practice of writing prompts, that is, the natural-language instructions given to a model to elicit a desired output, has evolved from an engineering curiosity into a widely discussed skill, variously labelled prompt engineering, AI literacy, or prompt literacy. This paper argues that prompting is best understood not merely as a technical or engineering competence but as a linguistic practice, one that draws on pragmatics, speech-act theory, register variation, and politeness strategies in ways that closely parallel ordinary human-to-human communication while also departing from it in theoretically interesting ways. Drawing on scholarship published between 2019 and 2026, spanning the original few-shot learning paradigm, systematic surveys of prompting techniques, Gricean and speech-act analyses of human-AI dialogue, and the emerging educational literature on prompt and AI literacy, the study develops a linguistic-functional taxonomy of prompting strategies and applies it, through qualitative content analysis, to a purposively selected corpus of widely documented prompt patterns, including zero-shot and few-shot instructions, chain-of-thought reasoning prompts, role or persona prompts, and structured academic frameworks such as the CLEAR model. The analysis finds that effective prompts systematically deploy directive speech acts, explicit and implicit context-setting, register calibration, and cooperative-principle-like maxims of quantity, quality, relation, and manner, adapted to the specific inferential behaviour of large language models. The study concludes that prompt engineering constitutes a genuine new literacy in the sociolinguistic sense, a describable, teachable set of language practices for achieving communicative goals with a non-human interlocutor, and it proposes an integrated pedagogical and linguistic research agenda for studying this literacy as it continues to develop.