Mathematical problem-solving skills are a key competency of the 21st century and are an important focus in mathematics education. This study aims to map the development of research on mathematical problem-solving skills in the period 2010–2025 using a bibliometric approach based on Google Scholar. Data were collected through Harzing's Publish or Perish (PoP) using keywords related to mathematical problem solving/problem solving ability and exported in .CSV and .RIS formats. The metadata were then cleaned and normalized using Microsoft Excel, then analyzed using VOSviewer to produce network, overlay, and density visualizations. The results of the analysis show an increasing publication trend since 2015, peaking in 2021–2022, then declining in 2023–2025, which is thought to be influenced by the limited coverage of current data. Co-occurrence mapping formed several main clusters: (1) learning models and problem-solving development, (2) skills/abilities as learning outcomes, (3) educational context and learning influences, (4) problem-solving processes and strategies as well as affective factors (self-efficacy), and (5) assessment and ability indicators. The overlay findings indicate that current topics for 2021–2025 are shifting toward affective factors and self-regulation, such as self-efficacy, mathematical resilience, independence, and the study of strategies and processes. Overall, research is still dominated by cognitive and pedagogical factors, while integrative exploration of affective aspects and the development of comprehensive assessment instruments remain limited. This mapping provides a basis for formulating directions for further research and strengthening the design of learning and assessment of mathematical problem-solving skills.