A randomized controlled trial in Sierra Leone found that students using Google DeepMind's Gemini Guided Learning feature showed measurable gains in both engagement and learning speed compared to control groups. This is not a pilot program or a press release. It is a published RCT, the methodological standard that separates signal from noise in educational research.
The study matters because it attempts to answer the question the entire edtech sector has been dodging: does AI tutoring actually move the needle on outcomes, or just on time-on-app metrics. The trial measured both, and the separation of those two variables is where the real argument lives. Read the methodology before you accept the headline.
The next question is replication. Sierra Leone is a specific context, with specific infrastructure constraints, specific student populations, and specific baseline literacy rates. Google DeepMind has a result. What they do not yet have is generalizability. That tension is worth your full attention.
[READ ORIGINAL →]