Building on Hilgard’s (1965) classic work, the domain of hypnosis has been conceptualised by Barnier, Dienes, and Mitchell (2008) as comprising three levels: (1) classic hypnotic items, (2) responding between and within items, and (3) state and trait. The current experiment investigates sense of agency across each of these three levels. Forty-six high hypnotisable participants completed an ideomotor (arm levitation), a challenge (arm rigidity) and a cognitive (anosmia) item either following a hypnotic induction (hypnosis condition) or without a hypnotic induction (wake condition). In a postexperimental inquiry, participants rated their feelings of control at three time points for each item: during the suggestion, test and cancellation phases. They also completed the Sense of Agency Rating Scale (Polito, Barnier, & Woody, 2013) for each item. Pass rates, control ratings, and agency scores fluctuated across the different types of items and for the three phases of each item; also, control ratings and agency scores often differed across participants who passed versus failed each item. Interestingly, whereas a hypnotic induction influenced the likelihood of passing items, it had no direct effect on agentive experiences. These results suggest that altered sense of agency is not a unidimensional or static quality “switched on” by hypnotic induction, but a dynamic multidimensional construct that varies across items, over time and according to whether individuals pass or fail suggestions.