The urgency is clear. A landmark population-based study found that autism spectrum disorder (ASD) affects approximately 2.64% of school-age children in South Korea, or about 1 in 38—one of the highest reported prevalence rates globally.¹ Unlike earlier estimates based solely on diagnosed cases, this figure emerged from active screening across entire school populations, revealing that a majority of children identified had never been previously diagnosed.¹

This gap between presence and diagnosis has become a defining challenge. In Korea, stigma surrounding developmental disorders has historically discouraged families from seeking formal evaluation, leading to underreporting in clinical systems. At the same time, the country’s robust national health databases have quietly accumulated vast amounts of behavioral, developmental, and medical data—creating the conditions for a different kind of solution.

Now, policymakers are turning to that data.

In 2025, South Korea launched a multi-million-dollar national initiative to apply big data and AI to autism screening, aiming to detect developmental differences earlier and more systematically across the population.² The project draws from integrated datasets including health insurance records, pediatric visits, developmental screenings, and potentially educational indicators—allowing algorithms to identify patterns that clinicians might miss.

The rationale is straightforward. Autism is typically diagnosable in early childhood, yet delays in identification remain common worldwide. Even in advanced health systems, many children are diagnosed years after symptoms first appear. In Korea, where cultural barriers can further delay evaluation, this lag can be especially pronounced.

AI offers a way around that bottleneck.

“Two-thirds of ASD cases… had not been previously diagnosed.” — National Institute of Mental Health summary of Kim et al. study¹

Instead of waiting for families to initiate evaluation, predictive models can flag risk signals based on routine data. Subtle indicators—missed developmental milestones, patterns in pediatric visits, speech delays recorded in medical notes—can be aggregated and analyzed at scale. In theory, this allows for population-level screening without requiring active participation at every step.

The implications are significant. Early intervention is widely associated with improved developmental outcomes, particularly in language and adaptive functioning. By shifting detection earlier, AI-driven systems could reduce the long-term burden on both families and public services.

South Korea’s approach is also shaped by its broader data ecosystem. The country’s National Health Insurance Service (NHIS) covers nearly the entire population, creating one of the most comprehensive health datasets in the world. This centralized system enables longitudinal tracking—following individuals over time, rather than relying on fragmented records.

That continuity matters. Recent analyses of NHIS-linked data show that ASD prevalence in Korean children has increased steadily from 2011 to 2021, with particularly sharp rises among younger age groups.³ Whether this reflects true increases, improved detection, or both remains an open question. But the trend reinforces the need for scalable screening systems.

Globally, autism prevalence is estimated at roughly 1% of children, though rates vary widely depending on methodology and access to screening.⁴ South Korea’s higher reported prevalence is often interpreted not as an outlier in incidence, but as evidence of more thorough identification—particularly when screening extends beyond clinically referred populations.

In that sense, Korea’s AI initiative represents a continuation of an earlier insight: that how a society looks for autism shapes what it finds.

Still, the shift toward algorithmic screening raises new questions.

One concern is transparency. Predictive models, especially those built on large administrative datasets, can be difficult to interpret. Families may be notified that a child is “at risk” without clear explanation of why. In a context where stigma remains a factor, the manner of communication may be as important as the detection itself.

There are also issues of equity. While Korea’s healthcare system is broadly accessible, disparities in follow-up care could persist. Identifying more children is only beneficial if diagnostic services and interventions scale alongside detection.

Privacy is another consideration. The same centralized data systems that enable AI screening also concentrate sensitive personal information. Ensuring that data is used responsibly—and with public trust—will be essential as these programs expand.

Despite these challenges, South Korea’s model is drawing international attention. As countries grapple with rising autism diagnoses and strained clinical systems, the idea of passive, data-driven screening is gaining traction.

What distinguishes Korea is not just its technology, but its willingness to apply it at a national scale.

The country’s experience suggests that autism may be less a hidden condition than a hidden dataset—one that becomes visible when systems are designed to look broadly enough. In the early 2010s, comprehensive screening revealed a prevalence far higher than expected. Today, AI is poised to extend that visibility even further, moving detection from the clinic into the background infrastructure of everyday life.

For policymakers and researchers, the lesson is both technical and cultural: data can close gaps in detection, but only if systems are built to overcome the social barriers that created those gaps in the first place.

As South Korea continues to refine its approach, it is not just advancing autism screening. It is redefining what it means to see.

¹ Young Shin Kim et al., “Prevalence of Autism Spectrum Disorders in a Total Population Sample,” American Journal of Psychiatry 168, no. 9 (2011): 904–912.
² South Korea national AI autism screening initiative, reported in Healthcare IT News (2025).
³ National Health Insurance Service (NHIS)–linked analyses of ASD prevalence trends in Korea, 2011–2021.
⁴ World Health Organization, global autism prevalence estimates (approximately 1%).