South Africa Middle Class Shifts Spending Toward AI

The joint SpendTrend 2026 analysis by Discovery and Visa shows middle-class South Africans are shifting subscription spend toward AI. Payment volume for AI subscriptions rose 125% in 2025 and 43% of Discovery Bank Visa clients in the sample now pay for an AI subscription. ChatGPT leads paid usage at 67%, Gemini at 35%, and Copilot at 27%, while weekly AI use for household shopping is 40%.

The SpendTrend dataset covers Discovery Bank clients who pay with Visa, a cohort that represents the upper middle class and skews toward higher-income consumers. Key usage metrics from the report include 40% weekly AI-assisted shopping, with sub-uses of 61% for price comparison, 53% for product research, and 47% for finding deals. Among AI-influenced purchases in the past 12 months, 42% switched to cheaper alternatives and 35% used AI to avoid risky purchases or scams. The report also highlights that consumers are actively managing subscriptions rather than passively accumulating them, with AI subscription share of total subscription payments doubling in 2025.

This is an early signal that paid consumer AI is moving beyond novelty toward transactional value, particularly in retail and personal finance. The dominance of ChatGPT underscores the advantage of broad conversational assistants for discovery and comparison tasks, while Gemini and Copilot show multi-vendor engagement. For product and data teams, the pattern suggests a shift from attention-driven monetization to utility-driven monetization, where measurable ROI in shopping and fraud avoidance justifies recurring payments. The upper-middle-class bias matters: this cohort is often first to adopt paid features, so these metrics are leading indicators rather than population averages.

Monitor whether AI subscription growth broadens beyond affluent cohorts and whether merchant referral traffic via AI, now 3.5% globally per cited data, scales to materially change affiliate and attribution models. Practitioners should track conversion lift from AI-driven recommendations, the evolution of privacy-consent flows for purchasing data, and how fraud-detection pipelines adapt as consumers rely on AI for risk screening.