Cognitive reframing, a core cognitive behavioral therapy (CBT) technique, has been widely recognized as an effective psychological intervention. However, clients often exhibit negative or resistant attitudes during treatment, posing a persistent challenge even for state-of-the-art large language models (LLMs) in cognitive reframing. To address this challenge, we propose the Cognitive Reframing Assistant with Reflective Engagement (CARE), which dynamically adapts to the user’s evolving state. Specifically, as a user’s cognition of an event evolves, their emotions change accordingly. By capturing these shifts, CARE helps users recognize and adjust negative thinking patterns, to alleviate emotional distress through structured multi-hop reasoning steps. Moreover, we introduce a novel metric, COG-DEBATE, to evaluate the effectiveness of cognitive reframing intuitively. Experimental results demonstrate that CARE outperforms baseline models, increasing the success rate of cognitive reframing by over 40%, effectively fostering independent thinking. Artificial experiments further show that our approach is highly effective in encouraging positive emotions (The code and data will be made public after the paper is accepted.).