Current large vision-language models (LVLMs) achieve remarkable progress. By combining pre-trained visual encoders with large language models, MLLMs can perform multimodal tasks such as describing images, visual question answering, etc. based on visual inputs and language instructions. Despite their remarkable achievements, LVLMs still suffer from hallucination, which is the problem of generating descriptions that include objects that do not actually exist in the images. Existing methods mitigate this issue with either training with specific designed data or inferencing with external knowledge from other sources, both incurring considerable additional costs. In this paper, we propose BOP: Briefness-Oriented Prompting elicits faithfulness in LVLM, a novel LVLM prompting method. Our approach begins with the observation that LVLM tends to generate more faithful response when prompted for a concise output. This phenomenon results in LVLM generating more hallucinations when asked to provide detailed image captions. Based on the observation, BOP dissects the initial prompt into fine-grained instructions. Through comprehensive experimentation across various LVLMs and metrics, we demonstrate BOP's substantial efficacy in mitigating hallucination, underscoring its effectiveness and versatility.