The first independent RCT of an AI dyslexia-intervention game found a real but modest gain for the lowest-performing US readers (ITT +0.095 SD, p≈.06), rising with dosage — though most students used it far below the recommended 9 hours.
848 students
Sample size (2022-2025)
+0.095 SD
Intent-to-treat effect
+0.20 SD
Dosage-response effect per 10x usage
+0.30 SD
Projected effect at recommended 9-hour dose
Details
Maturity
Scaling
Promoter
Dysolve AI (EduNational Solutions), independently evaluated by CRESP, University of Delaware
Period
2022–2025 school years (RCT); technical report published October 2025
Keywords
AI reading intervention, dyslexia, randomized controlled trial, learning disabilities, K-12
Context
Dysolve is a game-based AI program from EduNational Solutions that autonomously generates individually tailored reading games for struggling readers, without requiring teacher training. CRESP at the University of Delaware ran the first large-scale, fully independent RCT of the program.
Results
848 students in grades 3-8 across 32 schools in 9 US states, overwhelmingly low-income and minority and scoring near the 10th percentile at baseline, were randomly assigned within grade and school over the 2022-23 through 2024-25 school years. Intent-to-treat analysis found +0.095 SD (p≈.057); a dosage analysis found +0.20 SD per 10-fold increase in usage, projecting to roughly +0.30 SD at the recommended 9-hour dose.
Conclusions
Most treatment students used Dysolve well under the recommended dose, meaning the trial likely understates the programme's potential effect under full, faithful implementation, while also showing how fragile real-world impact is when dosage isn't met.
Implementation
Implementation detail (cost, timeline, staffing, conditions for success) is not yet available for this practice.
Data sources
Where this practice's information was retrieved from, and when.
Grammatek's neural Icelandic text-to-speech technology now narrates textbooks for Hljóðbókasafn Íslands, the national library serving dyslexic and print-disabled students, after …
A peer-reviewed Fiji Ministry of Education study combined child-functioning survey data with learning-support-needs data in a rule-based algorithm that more …
A solar-powered tablet game teaches literacy and numeracy to out-of-school children in conflict-affected Sudan. A 2020 peer-reviewed study found significantly …
