Abstract:

Generative AI (GenAI) literacy is emerging as a transversal professional skill, yet there is growing evidence that students preparing for knowledge-intensive careers remain underprepared to use these tools with the critical awareness, selectivity and integrity required in professional practice. This paper conceptualises GenAI literacy across five observable dimensions: adaptive prompting, post-response evaluation, tool selection, attribution and transparent disclosure, and integrity-oriented decision making, operationalised here based upon recent theorisations of GenAI literacy.

This paper maps observable literacy gaps across each dimension, drawing on mixed-methods data collected from engineering students across multiple cohorts and two European institutions, including survey data, qualitative video logs, thesis attribution declarations and raw prompting data. Findings include near-total reliance on a single general-purpose large-language model chatbot regardless of task, minimal awareness of domain-specific tools, exclusively language-focused wrapper use, high usage paired with low trust in GenAI outputs and superficial engagement with attribution requirements.

In response, the paper presents a bifurcated prompting framework that integrates pre-response prompt construction with post-response evaluation and iteration, embedding critical evaluation directly into the prompting structure and thereby operationalising two dimensions simultaneously. The paper argues that GenAI literacy does not develop spontaneously through tool exposure or digital nativeness but must be deliberately cultivated if emerging professionals are to participate effectively and with integrity in AI-mediated knowledge work.

Keywords:

GenAI literacy, transversal skills, academic integrity, AI prompting.