This work explores real-world deniability in messaging. We propose a formal model that considers the entire messaging system to analyze deniability in practice. Applying this model to the Signal application and DKIM-protected email, we demonstrate that these systems do not offer practical deniability guarantees. Additionally, we analyze 140 court cases in Switzerland that use conversations on messaging applications as evidence and find that none consider deniability, providing evidence that this property does not have an impact in the legal setting. Based on these technical and legal findings, we assess whether deniability is a desirable property and the challenges and shortcomings of designing a system that is deniable in practice. We posit that systems should either offer real-world deniability or refrain from claiming to achieve it. We discuss how to choose an appropriate threat model for deniability in a given context and how to design communication systems that are deniable in practice. For Signal, we propose and discuss a simple yet effective solution: the application should enable direct modification of locally stored messages in the user interface. This position paper raises several unanswered questions, aiming to further stimulate discussion and research on real-world deniability in messaging.