A vehicle remote instruction system includes a remote instruction point situation recognition unit that recognizes a remote instruction point situation on a target route based on the target route, position information, and map information, a recommended time acquisition unit that acquires a recommended time based on the remote instruction point situation and a type of a remote instruction, a remote instruction sending unit that sends the remote instruction to an autonomous driving vehicle, and a predicted time calculation unit that calculates a predicted time. The remote instruction sending unit sends a delay instruction to set a reflection time closer to the recommended time when the predicted time is earlier than the recommended time, and sends a preparation instruction to cause the autonomous driving vehicle to behave in response to the type of the remote instruction at an earlier time when the predicted time is later than the recommended time.