This research is motivated by the problems that researchers find in the field low interest and knowledge of students in learning to write descriptive texts. This is due to the lack of application of a unique and interesting learning model, so that it can make it easier for students to learn activities. This study aims to prove the influence of the Prediction Guide learning model on learning to write descriptive texts at SMPN 228 Jakarta. This study uses a quantitative approach with a pseudo-experimental method of non-equivalent group design research design. The population in this study is SMPN 228 Jakarta with a total of classes VII A to VII H with research samples of classes VII A as the control class and VII B as an experimental class with a total of 36 students in each class. The data collection technique uses test instruments with the help of data processing using SPSS software version 27. The results of the study were obtained that the average score of the experimental class was 66 pretest while 81 posttest. This proves that there is an increase in the results of scores in the experimental class where the Prediction Guide learning model is applied. The normality test as a prerequisite for the parametric test was then obtained abnormal data at 0.049 pretest of the control class and 0.042 posttest of the experimental class, while normal data at 0.050 posttest of the control class and 0.102 pretest of the experimental class. Then, the homogeneity test obtained a value of 0.977 for the control class and 0.819 for the experimental class, the data was declared homogeneous. The Mann-Whitney test obtained a significance value 0.003 < 0.05 until H_1 is accepted H_0 rejected. The N-Gain test obtained a value of 0.5232 with the medium category. Therefore, it can be concluded that the application of the Prediction Guide learning model has an effect on learning to write descriptive texts. .