The World Wide Web has become a popular source for gathering information and\nnews. Multimodal information, e.g., enriching text with photos, is typically\nused to convey the news more effectively or to attract attention. Photo content\ncan range from decorative, depict additional important information, or can even\ncontain misleading information. Therefore, automatic approaches to quantify\ncross-modal consistency of entity representation can support human assessors to\nevaluate the overall multimodal message, for instance, with regard to bias or\nsentiment. In some cases such measures could give hints to detect fake news,\nwhich is an increasingly important topic in today's society. In this paper, we\nintroduce a novel task of cross-modal consistency verification in real-world\nnews and present a multimodal approach to quantify the entity coherence between\nimage and text. Named entity linking is applied to extract persons, locations,\nand events from news texts. Several measures are suggested to calculate\ncross-modal similarity for these entities using state of the art approaches. In\ncontrast to previous work, our system automatically gathers example data from\nthe Web and is applicable to real-world news. Results on two novel datasets\nthat cover different languages, topics, and domains demonstrate the feasibility\nof our approach. Datasets and code are publicly available to foster research\ntowards this new direction.\n