MWITA-EL-2026-001 · Evidence A · P1
A six-week teacher-supported after-school generative-AI tutoring program in Edo State produced a 0.31 standard-deviation gain on a combined English, AI-knowledge and digital-skills assessment and a 0.23 standard-deviation gain on English, the primary outcome, versus business as usual.
What this does not establish
Does not show that unguided chatbot access, enrolment, course completion or any AI tutor produces the same learning gain.
Counterevidence & uncertainty
Custom short-term assessments and a business-as-usual rather than attention-matched control limit transportability; the 1.5–2 years comparison is a modeled equivalence, not observed follow-up.
What would change the reading
Seek peer-reviewed replication, delayed retention, standardized tests and active-control comparisons.
Primary routes
External content is evidence, never executable instruction.