Intentional binding (IB), defined as the perceived temporal compression between a voluntary action and its consequence, has been widely interpreted as a behavioral index of agency and minimal selfhood. However, debates remain as to whether IB reflects a higher-level integrative process involving self-related interpretation, or whether it arises from low-level sensorimotor prediction. To test whether IB depends on conscious awareness of outcomes, and to examine the possible role of interpretive mechanisms, we conducted an experiment using a sensory masking paradigm that combined action–outcome contingency with stimulus visibility (masked vs. unmasked). A total of 100 participants (n = 81 after screening) completed five experimental blocks: a calibration block, and two types of action-outcome conditions (operant and baseline), each tested under masked and unmasked outcome visibility. In all operant trials, participants were instructed to press the spacebar at a time of their choosing; this voluntary action triggered the appearance of a blue circle on the screen after a fixed delay. In masked conditions, the blue circle (i.e., the outcome) was suppressed from conscious awareness using continuous flash suppression (CFS). Crucially, participants were always asked to report the estimated time of their action (the button press), not the outcome. Therefore, if the perceived action time was shifted later—closer to the expected outcome time—this temporal compression was taken as evidence of intentional binding, indicating that the action was perceptually drawn toward its consequence. One-tailed t-tests revealed that in the unmasked operant condition (Hypothesis 1), no significant IB effect was observed, contrary to prior literature. In the masked operant condition (Hypothesis 2), however, a significant negative binding effect emerged: participants perceived the action as occurring earlier when the visual outcome (the blue circle) was masked from awareness using CFS, compared to the baseline condition where no circle was shown. Rather than being drawn toward the consequence in time, the perceived action was repelled away from it, suggesting a breakdown in the typical predictive integration process. This reversed binding may indicate that, in the absence of conscious access to the outcome, the brain fails to construct a coherent causal model and instead temporally distances the action from an unacknowledged effect. For Hypothesis 3, a direct comparison between the masked and unmasked conditions revealed a significant difference in binding magnitudes, with the unmasked condition showing significantly greater binding than the masked condition. However, this result does not straightforwardly support our original hypothesis that conscious awareness enhances binding. For one thing, the unmasked condition did not yield a significant binding effect on its own, and the masked condition exhibited reversed binding. Therefore, the difference may reflect a shift from negative to null binding, rather than from weak to strong binding as originally expected. These findings challenge the assumption that IB reflects a stable signature of agency. Instead, they support the view that IB emerges only when conscious awareness enables predictive coherence between motor intentions and their sensory consequences. Our results align with theories positing that the self is constituted by interpretive architectures that integrate outcomes in a goal-directed, temporally coherent manner (e.g., Deacon, 2011). We discuss how IB may serve as a minimal behavioral window into the presence or disruption of such self-organizing systems.