The game of poker has long served as a testbed in which AI agents are developed in order to create solutions for decision-making processes when dealing with incomplete information. In this thesis we perform a scientific literature review (SLR) to analyze how AI agents overcome uncertainty in Texas Hold'em Poker. The review includes peer-reviewed publications from 2011 to 2025. Initially, the number of papers was around 900, after applying inclusion and exclusion criteria this number was reduced to 120 relevant works for the detailed analysis. The focus lies on the agents that approximate an Īµ-Nash equilibrium, by computing game-theoretic strategies, utilizing abstraction techniques, and variants of Counterfactual Regret Minimization (CFR). We address these key research questions: (1) Which metrics are used to compare poker AIs? (2) What forms of uncertainty are identified in poker AI, and how are they addressed in both theory and practice? Key findings include the advancements in Counterfactual Regret Minimization (CFR) and its variants, combined with techniques to abstract the game in order to solve it in a realistic time manner. Under the pretense of uncertainty, these findings are evaluated and presented accordingly. We also highlight a paradigm shift that moves away from static abstraction to dynamic real-time solving, which enables the AIs to perform better against humans.