Sentiment analysis has been one of the hot topics for researchers for previous two decades. Researchers from domains like natural language processing (NLP), statistic, computational linguistics and information retrieval (IR) have been targeting different types of problems related to sentiment analysis. While the number of research problems related to sentiment analysis has been varying according to the global user requirements, major task of sentiment analysis has always been focus of the core sentiment analysis research i.e. to classify the given text into positive or negative categories (and sometimes a neutral category is also added). Researchers have been using several kinds of approaches for this core tasks while exploiting several types of features. Most of the time, researchers tend to use textual features for sentiment analysis task. While several have tried to exploit richness of non-textual features, focus has been revolving around the non-textual features like exploiting document structure, text position within the text, using social network evidences etc depending upon the nature of data collection being used. In this paper, we propose a novel approach of using credibility of sentiment analysis. Two twitter data sets are used for experimentation purposes collected over two different durations from Twitter using Twitter APIs. Both data collections consist of geotagged tweets from six different cities of Pakistan. The standard method of pooling is used for creating the gold standard. We exploit the credibility of information as strong evidence for sentiment analysis and the proposed approach has shown effectiveness in its results. A unique approach (“trust the trusted”) has been proposed to exploit the credibility of information for sentiment analysis purpose.