We developed a coupon-giving robot system for a shopping mall to explore possible applications using social robots in daily environments, particularly for advertising. The system provided information through conversations with people. The robot was semi-autonomous, which means that it was partly controlled by a human operator, to cope with the difficulty of speech recognition in real environments. We conducted two field trials to investigate two kinds of effectiveness related to recommendations: the presence of a robot and different conversation schemas. Although a robot can strongly attract people with its presence and interaction, it remains unknown whether it can increase the effects of advertisements in real environments. Our field trial results show that a small robot increased the number of people who printed coupons more than a normal-sized robot. The number of people who printed coupons also increased when the robot asked visitors to freely select from all coupon candidates or to listen to its recommendation.