<![CDATA[Both optimists and skeptics regarding the adoption of AI in primary education require clear evidence. The purpose of this study is to compare the contributions and weaknesses of artificial intelligence in primary schools. The method used is a systematic literature review with a light bibliometric analysis of articles from the period 2020–2025. The review focused on differentiated learning, formative assessment and feedback, explainable artificial intelligence, human loop mechanisms, and data governance. After gradual screening in accordance with PRISMA guidelines, 72 articles were analyzed thematically and mapped using keyword co-occurrence. The results show that most studies report improvements in literacy and numeracy when artificial intelligence is used as an educational strategy enhancer. With an average impact at a moderate level, the results show improvements in literacy and numeracy. Automatic feedback systems reduced teachers' routine workload by approximately 24 minutes per week and increased the frequency and accuracy of responses. While misconception detection was considered adequate, the level of agreement between artificial intelligence and teacher assessment was rated as moderate to high. However, fairness audits are uneven, teacher control features are limited, system explanations are lacking, and scientific openness is low. Research shows that artificial intelligence is most effective when it reinforces the pedagogical role of teachers and enforces students' intellectual responsibility. This can have an impact on policies regarding child-friendly explanations, strict data governance, and effective human procedures.]]>
