Large language models exhibit status quo bias in climate-relevant advice

Abstract

Large language models (LLMs) are increasingly used by consumers and decision-makers to provide advice, including on choices relevant to climate change. Status quo bias is the tendency to favor existing conditions over change. It is well-documented in human psychology and could be problematic in LLM climate advice given that rapid societal changes are required to sufficiently reduce emissions. We evaluated 7548 queries spanning multiple domains including vehicle purchases, recipes, home heating, and climate-relevant policy trade-offs. Each of these queries was tested on at least six different LLMs for a total of 54 888 prompts. By asking climate-relevant trade-off questions that held all variables constant except status quo framing, we found that LLMs had, on average, 7.5 times higher odds of recommending a course of action when it was framed as preserving existing conditions than when it required change, a finding that persists in leading 2026 models. In consumer contexts, LLMs were 4.1 times more likely to recommend electric vehicles to users who already owned one and vehicle recommendations correlated with regional adoption rates (rs = 0.48, p < .001) but rarely exceeded them. Recipe recommendations demonstrated no partiality for (low-emissions) plant-based options (40% vegetarian or vegan). These consumer findings complement the evidence from the more controlled tradeoff findings and together suggest that, on balance, LLMs systematically favor existing conditions in their recommendations. While conservative advice may be appropriate in many contexts, climate mitigation requires departure from high-emission norms. Users seeking climate-relevant advice from LLMs should be aware of this tendency, and developers of climate or consumer-oriented applications may wish to address it through system prompts.