Audience research is such a hit nowadays, and always has been based on web logs modeling. However, the content data of articles has not been considered yet. So, this paper clusters the ordered word vectors of an article to analyze audiences browsing habits of Chinese campus. This paper first obtains the contents and page views of articles from WeChat official accounts of several Chinese campuses. Then, it analyzes the audiences by clustering the ordered word vectors of the articles they have been read. To do so, the articles are first converted into a word vector model through Jieba word segmentation and word2vec, and then TF-IDF is applied for weight sequencing. Finally, the distances between articles are obtained by calculating the mean sum of square of the distances between words. Also, we cluster contents according to their topics and obtain many meaningful results. This paper provides a new method of audience analysis based on semantic, which can simplify the manual classification process of articles, and provide more comprehensive analysis conclusions. Moreover, these analysis results are useful for the strategy of content production and precious marketing.