在互联网技术迅速发展的背景下，在线客户服务互动中的多模态信息正变得越来越重要。保险机构在市场竞争和利润缩减的压力下，寻求通过数字化转型降低成本并推动业务增长，因此线上互动在保险公司的客户服务策略中变得至关重要。尽管如此，关于保险行业线上互动特别是多模态信息方面的研究仍然相对有限。本文通过分析众安保险公司在企业微信平台上的客户服务互动数据，探讨了多模态信息对客户回复和购买行为的影响。 本文构建了基于文本、视觉、听觉特征的分析模型，探讨多模态信息对客户行为的整体影响。文本信息特征方面，文本长度与客户回复和购买行为呈显著的负向关系，动词的使用比例显著正向影响客户的回复和购买行为，而量词使用对于客户回复和购买行为的影响存在差异。视觉信息特征（表情符号和图片的使用）与客户的回复和购买行为之间呈现出显著的正向关系。听觉信息特征中，音频信息的使用对客户回复和购买行为均有显著正向影响，但因模拟口语停顿和语调变化而被视为听觉线索的标点符号对客户回复行为与购买行为的影响不一致。 本文进一步探讨了视觉信息特征通过顾客感知价值影响客户行为的机制。结果表明，视觉信息通过增加购物乐趣和促进客户参与度，提升了顾客的感知价值，从而正向影响客户的回复和购买行为。但当通过满足客户需求来增加顾客感知价值时，客户的回复行为呈现出负向趋势，表明客户需求得到满足后，进一步沟通的意愿可能降低，客户的购买仍受到积极的影响。 此外，本文研究了客户关系时长对视觉信息特征的调节作用。结果显示，客户关系较短时，视觉信息对客户的回复和购买行为的正向影响更为明显，而随着关系时长的增加，视觉信息的正向影响会减弱；不同主题对于视觉信息的作用效果存在差异，服务主题下视觉信息对购买行为呈正向影响，而在营销主题下则可能呈负向。 本研究不仅丰富了在线客户服务互动的理论，还为保险行业提供了多模态信息应用于客户服务和销售策略的实践建议。这些发现有助于保险公司更有效地设计和实施在线客户服务策略，提升客户满意度和忠诚度，促进业务增长。

In the context of rapid technological advancements, multimodal information in online customer service interactions is becoming increasingly significant. Insurance companies, under the pressure of market competition and shrinking profit margins, seek to reduce costs and drive business growth through digital transformation. As a result, online interaction has become crucial in their customer service strategies. However, research focusing on online interaction in the insurance industry, particularly on multimodal information, remains relatively limited. This paper explores the impact of multimodal information on customer response and purchasing behavior by analyzing customer service interaction data from ZhongAn Insurance on the WeChat Enterprise platform.The study constructs an analysis model based on text, visual, and auditory features to investigate the overall impact of multimodal information on customer behavior. In terms of text features, text length is significantly and negatively correlated with customer response and purchasing behavior, while the proportion of verbs has a significant positive impact on both. The use of quantifiers shows varying impacts on customer response and purchasing behavior. Visual features (use of emoticons and images) are significantly positively correlated with customer response and purchasing behavior. In auditory features, the use of audio has a significant positive impact on both response and purchasing behavior, while punctuation marks, often viewed as auditory cues due to their simulation of speech pauses and intonation changes, show inconsistent effects on customer response and purchasing behavior.The paper further explores the mechanism by which visual information influences customer behavior through customer perceived value. The results indicate that visual information increases shopping pleasure and promotes customer engagement, enhancing customer perceived value, which in turn positively influences customer response and purchasing behavior. However, when customer needs are met to increase customer perceived value, response behavior tends to decline, suggesting that the willingness to communicate further may decrease even as the positive impact on purchasing continues.Additionally, the paper examines the moderating effect of customer relationship duration on visual information features. The results show that when the customer relationship is shorter, visual information has a more substantial positive impact on response and purchasing behavior, but as the relationship duration increases, this positive impact tends to weaken. The influence of visual information varies with different themes: it is positively associated with purchasing behavior under a service-oriented theme but may be negatively associated under a marketing-oriented theme.This research not only enriches the theory of online customer service interaction but also provides practical recommendations for the insurance industry on the application of multimodal information in customer service and sales strategies. These findings can help insurance companies design and implement more effective online customer service strategies to enhance customer satisfaction and loyalty, thereby promoting business growth.