The explosive growth of social networks, computing technologies, and natural language processing tools have enabled online marketing, brand promotions, and campaigns. In this regard, social media influencers have received great attention due to their excessive audience reach. However, automatic identification and analysis of such users face challenges of noisy text and computation which are further exacerbated for multilingual content. In this paper, we propose a novel method for the identification of micro-influencers and examine the evolution of such users in the digital landscape of Pakistan. First, we create labelled datasets for the classification of hashtags and micro-influencers using machine learning methods. Next, we build and classify a dataset of 14.4 million tweets into political and non-political categories to identify political micro-influencers. In addition, we present an analysis on the hashtag usage and location of micro-influencers. Moreover, we perform a longitudinal analysis of micro-influencers by capturing two snapshots. The analysis shows that the micro-influencers exhibit highly dynamic behaviour as 40% accounts of micro-influencers from 2018 does not exist in 2021. The location analysis revealed 37% micro-influencers are not from Pakistan. This research is useful for journalists, political scientists, and social media analysts.