CoT prompting does not always help inductive reasoning in LLMs—we show it can amplify errors and propose structured interventions to address this.