Blinding is not a single yes-or-no property of a trial. It is a set of role-specific safeguards, and each role, if unblinded, opens a different route for bias.
Participants who know their assignment can change their behavior — adherence, diet, activity, use of other care — in ways that differ between arms. This is performance bias: the groups no longer receive the same co-interventions, so the comparison of outcomes reflects the behavior change as well as the treatment.
Providers who know the assignment can treat the two arms differently in ways that are not part of the protocol: extra monitoring, additional medications, different encouragement. This is also performance bias, and it operates through the care delivered rather than through the participant's own choices.
Outcome assessors who know the assignment can measure or judge outcomes differently between arms — a slightly more attentive examination, a more generous reading of an ambiguous scan, a different threshold for calling an event. This is detection bias, and it is especially important for subjective outcomes such as pain scores or functional status, where judgment enters the measurement. For an objective outcome such as all-cause mortality, the assessor's knowledge has little room to change the recorded result.
A trial can be double-blind for participants and providers but have unblinded outcome assessors, or vice versa. The appraisal question is therefore not "was the trial blinded?" but "which roles were blinded, and for which outcomes does that matter?"