To probe how non-human primates (NHPs) decode temporal dynamic stimuli, we used a two-alternative forced choice task (2AFC), where the cue was dynamic: a movie snippet drawn from an animation that transforms one image into another. When the cue was drawn from either the beginning or end of the animation, thus heavily weighted towards one (the target) of both images (the choice pair), then primates performed at high levels of accuracy. For a subset of trials, however, the cue was ambiguous, drawn from the middle of the animation, containing information that could be associated to either image. Those trials, rewarded randomly and independent of choice, offered an opportunity to study the strategy the animals used trying to decode the cue. Despite being ambiguous, the primates exhibited a clear strategy, suggesting they were not aware that reward was given non-differentially. More specifically, they relied more on information provided at the end than at the beginning of those cues, consistent with the recency effect reported by numerous serial position studies. Interestingly and counterintuitively, this effect became stronger for sessions where the primates were already familiar with the stimuli. In other words, despite having rehearsed with the same stimuli in a previous session, the animals relied even more on a decision strategy that did not yield any benefits during a previous session. In the discussion section we speculate on what might cause this behavioral shift towards stronger bias, as well as why this behavior shows similarities with a repetition bias in humans known as the illusory truth effect.