Abstract Argumentation frameworks (AAFs) unify many approaches to computational argumentation. The dynamics of AAFs has been recently recognized as a potential power of AAFs to cope with dynamic settings such as argumentative dialogues. In this paper, we focus on an arguably understudied form of argumentation dynamics, referred to as preference-based extension enforcement, which constrains the commonly studied extension enforcement form by considering only attack removals that can be rationalized by some argument preference relation. In particular, we formalize preference-based extension enforcement and study its relationships with agent’s behaviourally revealed preferences. Finally, leveraging a computational framework for revealed preference argumentation in our earlier work, we develop sound and complete algorithms for many instances of preference-based extension enforcement.