AI for Special Education Teachers: What Helps and What the Research Says

Why Special Education Is Different

Special education teachers plan for students whose needs differ from each other, not just from the class average. One classroom may include students with dyslexia, autism, ADHD, cerebral palsy, or visual and hearing impairments. On top of teaching, there are individualised education plans (IEPs) to write, progress reports to prepare and families to keep informed.

AI tools can take some of this load, particularly the drafting and adapting of materials. They cannot replace the knowledge a teacher has of each student. This guide covers what the research says so far, where AI helps in practice, and the risks that deserve more attention than they usually get.

What the Research Says So Far

The evidence is growing, but it is earlier and more mixed than many blog posts suggest. A few recent studies are worth knowing:

• IEP goals written with AI. Waterfield and colleagues (2026) compared IEP goals written by teachers with and without ChatGPT and found no statistically significant difference in quality. A larger replication by Randolph and colleagues (2026, 111 participants) found a small advantage for AI-assisted goals in some analyses, but it was not consistent across statistical methods. The reasonable reading: AI can help draft goals without lowering quality, but it does not reliably make them better.
• Adaptive learning systems. A 2025 study by Liu described a reinforcement learning system that chose teaching strategies with 89% accuracy and handled rare situations far better than simpler models. It is promising, but it was tested by simulating interventions on an existing dataset, not with real students in classrooms.
• How often teachers actually use it. A 2026 survey by Naatz and Ruppar found that fewer than 15% of special education teachers used generative AI weekly and fewer than 5% used it daily. Most are still experimenting.

In short, AI is a useful assistant for specific tasks today. Claims that it transforms outcomes for students with disabilities are ahead of the evidence.

Practical Uses by Type of Need

• Dyslexia: Immersive Reader in Microsoft 365 lets students adjust font size and line spacing, hear text read aloud and see words split into syllables. Reading Progress helps teachers track reading fluency over time.
• Autism and ADHD: generative AI can draft social stories, visual schedules and short, step-by-step instructions tailored to a student's interests. These support the work of teachers and therapists; they do not replace it.
• Visual, hearing or speech needs: live captions in Google Meet and Microsoft Teams, screen readers, and features such as VoiceOver and Live Speech on iPad. Augmentative and alternative communication (AAC) apps increasingly use AI for word prediction and clearer speech output.

Many of these accessibility features are already included in school licences for Microsoft 365 and Google Workspace for Education, so check what you have before buying anything new.

Using AI to Draft IEPs Safely

IEP drafting is where teachers report the biggest time savings. A workflow that keeps the teacher in control:

1. Write a short, anonymised profile: age range, current levels, strengths and one or two priority areas. Do not include the student's name, diagnosis details or medical information.
2. Ask for draft goals in a specific format, for example: "Write three measurable reading goals for a 9-year-old who reads at a 6-year-old level. Use the format: condition, behaviour, criterion, timeframe."
3. Check each goal against your own assessment data. Rewrite anything that does not match what you know about the student.
4. Discuss the draft with colleagues and the family before it becomes part of the plan.

Creating Differentiated Materials

Generative AI is good at producing several versions of the same material. Example prompts:

• "Rewrite this science passage about plants for a student reading at a second-grade level. Use short sentences, one idea per sentence, and suggest where a picture would help."
• "Create a maths worksheet on adding fractions with visual fraction bars for a student with dyslexia. Use a clear font, plenty of space and no more than six questions."
• "Write a social story, in simple first-person sentences, about what happens during a fire drill at school."

Always read the output as if you were the student. Materials that are slightly wrong or confusing can cause more harm for a student with additional needs than for a typical learner.

Risks That Deserve Attention

Privacy

Data about students with disabilities is especially sensitive. Keep names, diagnoses and reports out of public AI tools, read the terms of any platform your school adopts, and follow your school's data protection rules.

Over-reliance

AI does not know a student's history, family situation or what happened at home this morning. Treat every suggestion as a draft that needs your professional judgement.

Accuracy

Generative AI can produce confident, plausible content that is wrong. Check facts, and check that strategies it suggests are appropriate for the student in front of you.

Equity and language

Schools with better devices, internet access and training benefit more, which can widen existing gaps. Many tools also work best in English, so teachers working in other languages may need extra time to check and adjust output.

Getting Started

• Start from a student need, not a tool. Pick one real problem, such as preparing reading material at three levels.
• Try one tool for a month before adding another.
• Tell families when you use tools that process any student information, and explain why.
• Share what works with colleagues, including what did not work.
• Keep watching progress through your own observation and assessment, not only the data a tool reports.

Used this way, AI gives special education teachers back some time for the part of the job that matters most: working directly with students.

This article is also available in Malay. Use the BM | EN switch at the top of the page.

References

1. Liu, H. (2025). Personalized learning support system for special education: a real-time feedback mechanism based on deep reinforcement learning. Frontiers in Psychology. doi.org/10.3389/fpsyg.2025.1658698
2. Naatz, A. J., & Ruppar, A. (2026). Special education teachers' use of generative artificial intelligence (AI): An exploratory survey of frequency and factors influencing adoption. Journal of Special Education Technology. doi.org/10.1177/01626434251379800
3. Randolph, K. M., Quinn, S. E., Sears, J. A., Howorth, S. K., Howerter, C. S., & Merry, K. E. (2026). Replicating and expanding the use of artificial intelligence to support special education practice: a mixed-methods investigation. Frontiers in Education. doi.org/10.3389/feduc.2026.1916444
4. Waterfield, D. A., Coleman, O. F., Welker, N. P., Kennedy, M. J., McDonald, S. D., & Cook, B. G. (2026). IEPs in the age of AI: Examining IEP goals written with and without ChatGPT. Journal of Special Education Technology. doi.org/10.1177/01626434251324592