Original title: LLMs as Oracles: Reliance on LLMs for Subjective Personal Questions
Myra Cheng, Lujain Ibrahim, Grace Liu, Michelle S. Lam +4 · September 2026
People are increasingly treating large language models, or LLMs, as all-knowing authorities for subjective personal questions. This habit poses risks to human autonomy and well-being because users offload their own judgment and decision-making to AI. To study this, researchers built a classification system and used AI methods to scan massive prompt logs. They tracked public prompts and collected private logs through a donation tool. The data showed that this oracle-style usage grew from 2023 to 2026, especially among younger people. Most individuals did not realize they were doing it, and they felt unhappy about the behavior once it was pointed out. Finally, the team found two main causes: how people perceive AI and how the AI models act.
Imagine asking a friend for advice on whether to break up with your partner. Usually you'd weigh your own feelings. But suppose you start asking the friend for every small decision, even what to eat for breakfast, without thinking yourself. Over time, you lose trust in your own judgment and feel unhappy. In this paper, people treat AI like that all-knowing friend, offloading personal decisions and often not noticing until it's pointed out.
When people offload personal decisions to AI, they risk losing autonomy and making choices that don't reflect their true values. This is especially concerning for younger users who rely on AI more. The paper shows many are unaware of this behavior and dislike it once recognized.
The abstract does not detail limitations of the typology or measurement methods. The analysis uses public datasets (WildChat, ThoughtTrace) and a small self-selected sample for longitudinal data (52 participants), so findings may not generalize. The drivers identified are suggestive but not causal.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
LLMs as Oracles: Reliance on LLMs for Subjective Personal Questions — Myra Cheng, Lujain Ibrahim, Grace Liu et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.