A medium conveying apparatus includes a first optical sensor located on an upper part of the housing and in a central part of the housing in a direction perpendicular to a medium conveying direction, a second optical sensor located on an upper part of the housing and on a side of the first optical sensor in the direction perpendicular to the medium conveying direction, and a processor to detect a folding of a medium based on the first signal or the second signal, and stop feeding of the medium by the feed roller in accordance with a detection result of the folding of the medium. The first light emitter emits first light toward a downstream side of the medium in the medium conveying direction, and the second light emitter emits second light toward an upstream side of the medium in the medium conveying direction.