A key method for understanding the evolution of languages is to look for words which share etymological roots across languages. These words, called lexical cognates, allow historical linguists to group languages into families and study their structural history. Existing research on cognate clustering is chiefly based on analyses performed on lexical databases which limits the scope for exploring new cognates. Additionally, while the Indo-European family is a transcontinental one, its cognate analyses are largely limited to a number of European languages. In this research work we build a new dataset using word embeddings to search for cognates using phonetic matching between translations of words in different languages across the Indo-European family. The context-dependent positioning of words in word embeddings allows for comparisons with contextually similar words and hence clustering using unsupervised learning algorithms. We successfully find significant distinctions between these “contextually close clusters of cognates” among languages across the Indo-Iranian, Romance, and Germanic families. We release this novel method to demonstrate which words a language is more likely to lend to another.