Artificial Intelligence (AI) tutoring systems promise to personalize learning, enhance comprehension, and democratize education. Yet concerns persist regarding their potential to exacerbate existing inequalities. This study investigates the social impact of AI tutoring across socioeconomic and geographic divides in India. A survey of 150 secondary and higher-secondary students assessed access, learning experiences, equity concerns, and perceived social outcomes. Results show that AI tutoring improves comprehension (M=4.1) and exam confidence (M=3.9) but significant disparities remain. Rural students reported greater accessibility challenges and higher perceived inequality (Equity_ Mean = 3.51 vs 3.65, Urban). Statistical tests revealed significant differences in social impact between rural and urban students (t=-2.051, p=0.043). Income-based comparisons showed marginal trends: high-income students benefited more in learning outcomes, while low-income groups reported higher social gains. ANOVA across income brackets revealed no statistically significant group differences. These findings highlight AI tutoring dual role: while enhancing individual learning, it risks widening systemic divides unless supported by inclusive policies. Recommendations include subsidized access, offline-first platforms, and culturally adaptive AI content.