South Africa’s weak early reading rates put AI’s classroom value under the microscope

South Africa’s early-learning reading challenge has sharpened attention on what artificial intelligence (AI) can realistically offer classrooms, as educators and publishers consider whether digital tools can identify gaps, guide teaching and show whether learning improves.

How stark is the reading problem?

The Department of Basic Education’s Funda Uphumelele National Survey cited by the 2030 Reading Panel in 2026 found that 15% of learners at the end of Grade 3 cannot read a single word. The survey assessed 27,838 learners across 710 schools and reported that about three in 10 learners in Grades 1 to 3 met their home-language reading benchmarks.

From assessment to action: what educators want from AI

At Oxford University Press South Africa’s Future Learning Labs in Cape Town and Johannesburg in September, teachers, publishing specialists and researchers discussed how digital tools and AI might support teaching and learning. Participants linked AI’s potential to four practical classroom needs: identifying gaps in understanding, selecting appropriate support, assessing independent learner ability, and clarifying responsibility for content presented to children.

Oxford’s digital product manager and programme facilitator, Heather Brogan, said:

“We need to start by asking what a learner is struggling with and what support a teacher needs.”

Brogan emphasised that schools should first identify a specific educational difficulty and then consider whether a tool could help and how its contribution would be assessed.

Assessment must inform teaching

Angelique Timms, curriculum lead at Curro Holdings, put assessment at the centre of instructional decisions. She said:

“A reading assessment should help us decide what to teach next.”

Timms added that when teachers understand where a child is struggling, they can target support and check whether it is working, and that technology becomes useful when it helps that process.

The discussions noted that a completed assignment alone can give only a limited view of understanding. Where AI assists in producing an answer, teachers may need follow-up exchanges asking learners to describe their reasoning, explain evidence used, or apply knowledge to a different question to determine which parts of the work the learner can do without assistance.

Building AI literacy and classroom practices

Speakers suggested classroom activities that develop skills needed to evaluate AI output: vocabulary, comprehension and subject knowledge that help learners recognise incomplete explanations and ask deeper questions. Suggested activities included comparing an AI-generated answer with another source, identifying unsupported claims, and explaining why one account is more convincing than another. These practices, the programme argued, would let teachers assess the reasoning behind learners’ conclusions and make AI use part of a wider process of questioning and checking information.

Responsibility, oversight and context

Yolandi Farham, Oxford’s product director for Africa, said:

“Every decision to introduce AI carries a responsibility to the learner.”

She said those adopting AI must be clear about who checks content, how teachers guide its use, and how concerns are addressed.

Shirley Eadie of Human Studios and Lefa AI argued for locally informed approaches. She said:

“South Africa needs an approach informed by its own classrooms. Teachers, learners and parents should help identify the opportunities and the risks. Language, access and the realities of teaching must be part of that conversation.”

That approach, participants said, requires considering whether learners can access a tool, whether its language and content suit their needs, and whether teachers have the support and time to review generated material.

What success would look like

Speakers returned to a single practical measure: for a child approaching the end of Grade 3, the contribution of any intervention should be visible in learning through a growing ability to read, understand, explain and question. The value of digital tools, including AI, will depend on how well they help teachers target the skills learners still need and on clear arrangements for oversight and accountability.