This chapter examines the potentials and perils of artificial intelligence (AI) in moderating and mediating peace-related discourse in conflict-affected digital spaces. Focusing on the Turkish platform Ekşi Sözlük, the study introduces a five-role typology—Observer, Moderator, Interpreter, Facilitator, and Suppressor—to analyze how AI systems interact with vernacular expressions of political dissent, irony, and affect. Drawing on a hybrid empirical approach combining thematic coding of 150 platform entries and expert interviews with journalists, platform designers, and digital rights advocates, the chapter interrogates how algorithmic systems shape, distort, or suppress user-generated discourse in a polarized media ecosystem. The findings reveal that while AI can enhance transparency and support constructive engagement through empathy-oriented design (“peace prompts”), it often fails to interpret culturally embedded irony, flags satire as toxicity, and reinforces epistemic asymmetries. Experts emphasize the need for culturally sensitive reinforcement learning, civic auditing mechanisms, and UX-level interventions that respect discursive ambiguity and democratic discomfort. These insights are directly linked to the ambitions of Sustainable Development Goal 16, which seeks to promote just, peaceful, and inclusive societies. The chapter argues that for AI to contribute meaningfully to SDG 16, it must be understood not simply as a tool for content moderation, but as an infrastructural actor in discursive governance. By situating algorithmic peacebuilding within localized political cultures and platform practices, this chapter contributes a nuanced, context-sensitive framework for AI ethics in conflict-affected digital environments.