South Africa likes to see itself as a digital laggard, boxed in by infrastructure gaps, load shedding, and affordability ceilings. New OECD-Cisco research complicates that story. On paper, South Africans are among the world’s most enthusiastic adopters of generative AI. The harder question is whether that enthusiasm is improving daily life or simply stretching it thinner.
According to the OECD-Cisco Digital Well-being research, 45.4% of South African respondents say they actively use generative AI. That puts the country ahead of several wealthier economies, including Canada and South Korea. Trust levels are similarly high, with more than three-quarters of local respondents describing AI as useful. For a country often framed as catching up, those numbers land as an uncomfortable surprise.
They also flatten the story.
High adoption doesn’t exist in isolation. The same research shows that South Africa mirrors other fast-adopting emerging markets in less comfortable ways. Recreational screen time is high, digital-only socialising is common, and respondents report sharper emotional highs and lows linked to technology use. A significant share of South Africans say they spend more than five hours a day on screens for personal use, a level the OECD consistently associates with poorer wellbeing outcomes. The findings form part of the OECD’s broader Digital Well-being Hub research, which tracks how digital engagement shapes everyday life across countries and age groups: https://www.oecd.org/en/data/tools/digital-well-being-hub.html
That five-hour mark matters. OECD analysis shows that once personal screen time pushes past it, the odds of poor mental wellbeing rise sharply. People in that bracket are far more likely to report low life satisfaction and a weakened sense of purpose, even when sleep, income stress, and physical activity are accounted for. Below that point, moderate use often correlates with better outcomes. Above it, the trade-offs stack up quickly.
This isn’t a warning against technology or AI itself. The data actually complicates the usual screen-time moralising. People with no digital engagement often report worse wellbeing than moderate users. The problem isn’t digital tools. It’s imbalance. And imbalance is baked into platforms designed to keep attention engaged long after utility has peaked.
Generational divides sharpen the picture further. Under-35s are driving South Africa’s AI uptake, which mirrors the growing gap between how quickly AI tools are being adopted and how slowly skills and safeguards are catching up, enthusiasm outpacing preparation in ways already visible across the local tech sector.
Older South Africans are having a different experience. Usage drops sharply after 45, and uncertainty replaces enthusiasm. Among over-55s, many respondents say they don’t know whether they trust AI at all. That hesitation looks less like resistance and more like exclusion. In a country where digital divides already track age, income, and geography, uncertainty quickly becomes a structural disadvantage.
What really sharpens the wellbeing risk is loneliness. OECD modelling shows that heavy screen use becomes far more damaging when paired with social isolation or unemployment, dramatically increasing the likelihood of poor mental wellbeing. High screen time on its own is associated with worse outcomes. Combined with loneliness, the effect intensifies.
That combination is not rare in South Africa. It’s structural. High unemployment, long commutes, remote work, and digital-first socialising all push people towards screens that substitute for physical connection. AI tools don’t create that reality, but they fit neatly into it, often extending time online rather than reducing it.
The OECD also draws a useful distinction between screens used because you have to and screens used because you can’t switch off. Personal, discretionary screen time is far more predictive of poor wellbeing outcomes than time spent online for work. Generative AI increasingly blurs that line, sliding from productivity tool into always-on companion and quietly lengthening the digital day.
Cisco’s framing of the research leans heavily towards skills and access, pointing to training programmes and digital inclusion initiatives as the solution. Those matter, but they’re incomplete. South Africa doesn’t need convincing to adopt AI. That part is already happening. What’s missing is a serious conversation about limits, incentives, and the emotional cost of systems designed for constant engagement.
A more useful measure of AI success isn’t how quickly people start using it, but whether they can step away without penalty. For South Africa, the challenge now is less about adoption curves and more about whether enthusiasm turns into exhaustion.
Speed is easy. What comes next will show whether South Africa’s AI adoption is improving life, or simply demanding more of it.