AI interventions show gains for students with learning disabilities

Artificial intelligence (AI) offers unprecedented opportunities to make education more inclusive and effective. A new study assesses whether AI-based educational interventions genuinely improve learning outcomes for students with learning disabilities.

The study, The Effectiveness of Artificial Intelligence-Based Interventions for Students with Learning Disabilities: A Systematic Review, published in Brain Sciences (2025), examines how AI tools influence education for students with dyslexia, dyscalculia, and other specific learning disorders.

How effective are AI-based educational tools?

The review analyzed 11 experimental studies conducted between 2022 and 2025, covering 3033 students across primary, secondary, and university levels. All interventions used identifiable AI components, including personalized learning systems, intelligent tutoring, generative AI, and assistive applications.

The findings were strikingly consistent: every study reported positive outcomes. Students demonstrated measurable gains in areas such as reading comprehension, arithmetic fluency, attention, memory, and overall literacy. Some interventions showed substantial effect sizes, with improvements in arithmetic fluency reaching d = 1.63 and significant boosts in working memory and processing speed. For example, generative AI tools like ChatGPT delivered dramatic improvements in reading comprehension for children with dyslexia, while game-based learning increased digital literacy by more than 50%.

Despite the variations in study designs, the positive outcomes suggest AI has a transformative potential in special education. These technologies adapt to individual learning patterns, providing tailored support that traditional methods often cannot achieve. The review emphasizes that AI’s ability to deliver personalized feedback and reduce stigma associated with traditional learning support makes it particularly valuable for students who struggle in conventional classroom environments.

What challenges and risks are associated with AI in special education?

While the immediate outcomes are promising, the research highlights significant methodological weaknesses across the studies. None of the included experiments were rated as having a low risk of bias. Seventy percent were classified as having moderate risk, while the remaining 30 percent had high or serious risk. The primary issues were a lack of randomization, absence of control groups, and small sample sizes, which weaken the reliability of conclusions.

The review also raises concerns about publication bias. The uniform positivity of findings is statistically improbable, suggesting that studies with neutral or negative results may not have been published. This underreporting creates an overly optimistic view of AI interventions.

Another critical gap identified is the absence of long-term follow-up. None of the studies tracked participants over extended periods to assess whether learning improvements persisted or if reliance on AI led to negative consequences. The authors discuss the risk of cognitive offloading, where students become dependent on technology to perform tasks, potentially undermining the development of core cognitive skills. There is evidence that while AI may enhance procedural skills, it does not always support deeper conceptual learning. Without long-term data, it remains unclear whether AI interventions lead to lasting educational benefits or temporary performance boosts.

The authors also warn about the need for balanced integration of AI into educational systems. Overreliance on AI without adequate teacher training, ethical guidelines, and monitoring could inadvertently harm students. The technology must complement, not replace, skilled instruction.

What does this mean for educators and policymakers?

For educators, the review suggests that AI tools can significantly enhance learning when used as a supplement to high-quality teaching. Tools like text-to-speech applications, adaptive games, and AI-powered tutoring can discreetly support students, helping them build skills without the stigma sometimes attached to traditional learning support.

However, the authors caution against widespread implementation without further evidence. Policymakers are urged to fund pilot programs and support rigorous evaluations before mandating AI technologies in special education. Training teachers to use these tools effectively and ethically is also critical to maximizing their benefits.

For researchers, the review highlights clear directions for future work. High-quality randomized controlled trials with larger sample sizes are needed to establish a stronger evidence base. Longitudinal studies should be prioritized to determine whether learning gains are retained and to monitor for potential negative effects such as skill degradation from cognitive offloading. Furthermore, researchers are encouraged to include students’ own perspectives in future analyses, as their experiences provide valuable insights into how AI affects learning and self-advocacy.

The review also calls for the development of standardized outcome measures to enable meaningful comparisons across studies. Current heterogeneity in methods makes it difficult to synthesize findings or conduct robust meta-analyses. Finally, more transparency is needed, including the publication of null or negative results, to build a balanced understanding of AI’s role in special education.