Recent AI tools have become capable of learning about users and developing psychological profiles based on usage or publicly available data, such as social media content. This research examines how AI-driven disinformation campaigns can leverage psychological profiling and exploit cognitive biases to manipulate individual beliefs and decision-making. We introduce the AI Psychographic Algorithmic Targeting (AIPAT) Framework, which explains how large language models can be used to generate tailored disinformation targeting users based on their social media behavior. Through algorithmic targeting and the exploitation of cognitive biases, we explore the risks of AI-induced polarization and radicalization. This study contributes a theoretical foundation for understanding AI's role in disinformation and outlines a research agenda focused on developing effective mitigation strategies.