Many studies have shown that children with Autism Spectrum Disorders show lack of skills using social and communication skills. However, they understand, assimilate and retain better information using virtual aids. Therefore, the objective of this study has been to assess the effectiveness of an augmented reality training program based on visual support for students with Autism Spectrum Disorder to improve their social skills. The investigation was carried out following a quantitative approach, quasi-experimental method and pretest posttest design with control group. Participants, were 11 individuals (10 male and one female) grouped into control (n = 5) and experimental group (n = 6). They were assigned using a non-probabilistic intentional sampling method. Likewise, the experimental group worked with different augmented reality activities such as a player who had to score a goal; playing with a cow. In contrast, control group received a similar intervention but without the use of this tool; for instance pick up the objects according to the therapist's orders. The intervention lasted for 20 weeks, in 15 min sessions twice a week. The instrument for data collection was the Autistic Spectrum Inventory of Riviere. The Quicker Vision application was used as the AR-based intervention method. The results of the experiment do not show statistically significant differences between both groups, although slight improvements appear in some items. For example item 8 (flexibility) and item 11 (imitation). This indicates that in the future we could work on areas that have not been studied in other research. Nonetheless, it is necessary to develop a greater number of investigations in this aspect to obtain a better knowledge of its effects.