The cognitive footprint of large language models: User awareness and sustainable intentions

The rapid adoption of Large Language Models (LLMs) has raised concerns about their environmental impacts, yet limited research has examined how users understand and respond to these costs. This qualitative study explores the cognitive footprint of LLMs—defined as the psychological processes through which users internalize and respond to a technology’s environmental impacts, including awareness, interpretation, emotional response, cost-benefit reasoning, responsibility attribution, and perceived behavioral control (PBC). Guided by the Theory of Planned Behavior (TPB) and Value-Belief-Norm (VBN) Theory, semi-structured interviews were conducted with fifteen adult participants representing diverse professional backgrounds and levels of LLM engagement. Thematic analysis identified five core themes: (a) environmental impacts remained abstract until quantified through relatable comparisons, (b) emotional responses varied by technical background, (c) participants engaged in cost-benefit reasoning reflecting an individual-level sustainability paradox, (d) responsibility was externalized toward technology providers, and (e) sustainable intentions were conditional on low behavioral friction. Findings indicate that the awareness gap identified in prior research is not solely an information deficit but a framing deficit, and that PBC in opaque digital systems is structurally suppressed by a lack of system transparency and user-facing choice architecture. This study extends TPB and VBN into digitally mediated sustainability contexts, introduces the cognitive footprint as a measurable construct, and highlights the role of interface design in bridging the gap between environmental concern and sustainable AI usage.