When ChatGPT arrived, I was asked for my view on whether it could work as assistive technology for students with learning disabilities, and I gave an expert opinion for Dystinct (Ross, 2023). Since then the tools have moved fast, and the question has only become more pressing. So here is my considered take, held in the same evidence-led, cautiously optimistic spirit I bring to all technology.
The short answer is yes, AI can genuinely help dyslexic and neurodivergent learners. But how it helps, and what we need to watch, both deserve care.
Where AI genuinely helps
Much of what dyslexic learners find hard isn’t thinking. It’s the mechanics that sit between their ideas and the page. This is exactly where these tools can lift a barrier.
For writing, AI can act as a patient scaffold: helping a young person get started, organise a jumble of ideas into an order, or turn rough notes into a first draft they can then shape. For reading, it can summarise a dense passage or rephrase something in plainer language, so a text that felt like a wall becomes something a learner can get into. It never gets impatient, never sighs, and is available at 11pm the night before a deadline when no adult is. For a young person who has spent years feeling that written work is a fight, that can be quietly transformative.
The theme running through all of this is the one I care about most across my technology research: independence. Like the reading devices I’ve studied directly, AI can let a young person do things for themselves that previously needed an adult, and do them without their difficulty being on display.
Where we need to be careful
None of that means handing a child a chatbot and walking away. The cautions are real, and some are specific to how these tools work.
They get things wrong, confidently. AI can produce fluent, plausible text that is simply inaccurate, and a learner who struggles with reading may be less well placed to spot it. That makes adult guidance and critical-checking skills more important, not less.
There’s a line between support and substitution. Using AI to organise your ideas or check your draft builds skills. Using it to produce work you don’t engage with builds nothing. Where that line sits depends on the task and the young person, and it needs thinking about, not assuming.
Access isn’t equal. If AI becomes a serious support tool, the young people who can afford the good versions, and who have adults to guide them, pull ahead of those who can’t. That’s the same inequality of access I’ve documented elsewhere in dyslexia support, in a new form.
And data and safeguarding matter. These are children, and the privacy, safety and data implications of the tools they use need proper attention from schools, not an afterthought.
The honest position
AI is a powerful new addition to the assistive technology toolkit, and I’d rather schools engage with it thoughtfully than ban it and drive it underground. Used well, with adult guidance, clear boundaries and attention to accuracy and equity, it can remove real barriers for dyslexic learners and support their independence. Used carelessly, it can mislead, deskill or widen gaps.
That balance, of genuine opportunity held alongside honest caution, is exactly why independent, evidence-led evaluation of these tools matters so much, which is a subject I take up in evaluating EdTech honestly. The technology will keep changing. The principles for using it wisely won’t.
If your school or organisation wants help thinking through AI and assistive technology for neurodivergent learners, that’s part of my research and consultancy work.
Questions people often ask
Is ChatGPT good or bad for dyslexic students? Neither on its own. Used with guidance, it can be a genuinely helpful support for reading and writing and can build independence. Used without boundaries or checking, it can mislead or replace the very skills a learner needs to develop.
Will using AI stop my child learning to write? Not if it’s used to scaffold rather than to substitute. Getting help to organise ideas or check a draft builds skills; handing over work you never engage with doesn’t. The distinction matters more than the tool.
Can AI-generated information be trusted? Not blindly. AI can state wrong things very convincingly, and a learner who finds reading hard may struggle to catch errors. Treat it as a helpful assistant to be checked, not an authority.
Should schools allow AI tools? I’d favour thoughtful engagement over outright bans, alongside clear guidance, attention to equity, and proper regard for children’s data and safety. Banning tends to drive use underground rather than making it safer.
References and further reading
- Ross, H. (2023) ChatGPT for students with dyslexia? Expert opinion on ChatGPT as an assistive technology tool for students with learning disabilities, Dystinct. Read the expert opinion
- Ross, H. (2024) Impact of OrCam Learn on reading comprehension, Support for Learning, 39(3), 95-103. https://doi.org/10.1111/1467-9604.12474
- Ross, H. (2020) Review 3: Technology-led interventions for SpLDs, Council for Science and Technology. Read the review
See the full list of publications.