International surveys repeatedly showed that Japanese children were good at mathematics but disliking it. We hypothesized there were a considerable number of "fake math-dislikes" among Japanese students who claimed they disliked mathematics explicitly but accepted it implicitly. To examine this hypothesis, we administered questionnaires and paper-based implicit association tests (IAT) to 204 Japanese junior high school students (13-14 years old) and found 38 fake math-dislikes (Study 1). We hypothesized further that those fake math-dislikes would become real math-dislikes eventually and that informing of their implicit attitude toward mathematics might work preventing this undesirable transition. Then, in Study 2, we randomly assigned them into experimental and control groups and informed only the experimental students of their positive implicit attitude toward mathematics we revealed with the IAT. One year later, we found 15 of the 16 experimental students improved their math achievement scores while only 9 of the 17 control students did. The simple practice of informing of their implicit attitude worked effectively for improving their math achievement. As hypothesized, it prevented the fake math-dislike students from turning into real math-dislikes.