Dedicated time did not produce meaningful use
The two randomized trials involved elementary students in districts serving high-poverty populations. Control students were expected to use the AI literacy platform independently during scheduled sessions, yet only 60.7 percent in one district and 53.3 percent in the other ever logged on.
Average weekly use in the control groups was 2.18 and 5.23 minutes. The provider said students generally needed at least 30 minutes a week for academic benefits, so the study never approached the dosage required to evaluate the tutor's learning potential.
The human layer improved engagement but did not rescue the outcome
Tutors did not deliver reading instruction. They used check-ins, reflection, motivation, group norms, and technical troubleshooting to help students engage with the platform. That support raised weekly use by about one minute in one district and 4.4 minutes in the other, while stories completed rose 71 and 80 percent relative to control averages.
Usage still remained low, and the researchers found no statistically significant improvement in reading achievement. Students who were already higher achieving and less likely to receive special-education services were also more likely to use the platform, raising the risk that opt-in technology can reproduce the participation gaps it is sold to close.
Go to the source
Read the evidence behind this analysis. External links open in a new tab.
Futurism — Students given access to an AI tutor did not meaningfully use it Stanford SCALE Initiative — Access Is Not Enough
