Part of: Assessment in the AI Era: Adapting the AI Assessment Scale for African Classrooms, a practical guide to AI-era assessment for schools and assessment teams.
Updated November 2026. The AI Assessment Scale has been revised since this post first appeared in March 2025. The section on the 2025 update below shows how the original levels map to the current ones.
When school leaders in South Africa ask me about AI, the first question is still usually about cheating. It is the wrong place to start. The more useful question is what each task is meant to show, and how much AI use still lets it show that. The AI Assessment Scale gives teachers a simple way to answer that question task by task, and with some adaptation it works in South African classrooms, from well resourced schools to those with no reliable internet.
This post explains the scale, how to adapt it to South African conditions, and how a school can start using it this term.
What is the AI Assessment Scale?
The AI Assessment Scale (AIAS) was developed by Mike Perkins, Jasper Roe, Jason MacVaugh and Leon Furze. Rather than treating AI as either permitted or banned, it gives teachers five levels for saying how AI may be used in a particular task.
That moves the conversation away from the “cheating” framing that has dominated since ChatGPT’s release in November 2022. Teachers decide what counts as misconduct for each task, and the scale gives them a shared language for telling learners clearly.
The five levels
These are the levels of the original scale, the version most South African schools I speak to have seen. The current names are in the next section.
Level 1: No AI
Learners complete the task entirely without generative AI. This suits supervised work, handwritten tasks and examinations where personal knowledge and skills must be shown. In South Africa it also protects fairness, because it removes the advantage some learners have from AI access at home.
Level 2: AI-assisted idea generation and structuring
Learners can use AI to brainstorm and structure ideas, but produce the final work themselves. This can help learners working in an additional language rather than their home language, as long as they develop and express the ideas in their own words.
Level 3: AI-assisted editing
Learners write the original work and may use AI to refine and edit it. They submit both the original draft and the edited version, so the teacher can see what the learner wrote and what AI changed.
Level 4: AI task completion, human evaluation
Learners use AI for specific parts of the task and critically evaluate what it produces. This builds an understanding of what AI does well and badly, which learners will need in a workplace shaped by these tools.
Level 5: Full AI
AI can be used throughout the task. This suits work where using AI is part of the learning outcome, for example a project in a well resourced school exploring AI as a collaborator.

What changed in the 2025 update
The authors have since revised the scale. The new version, set out in a 2025 journal article and on aiassessmentscale.com, treats the levels as different kinds of task rather than steps from less AI to more, and focuses on keeping assessment valid rather than policing AI use. The main changes are:
- Level 1, No AI stays the same, and is now the only level where supervised conditions are required.
- Level 2 becomes AI Planning: AI may support brainstorming, outlining and research.
- Level 3 becomes AI Collaboration: AI may help draft, give feedback and refine, with the learner evaluating and changing what it produces.
- The old Level 4 is removed, and Full AI moves to Level 4.
- Level 5 is new, AI Exploration: learners use AI creatively to solve problems or find new approaches in the subject.
If your school already uses the original levels, there is no need to start again. The principle is the same: the teacher sets the level, explains it, and designs the task so the evidence matches. When you next review your policy, the current version is the one to adopt. My guide to AI-era assessment for African classrooms uses the current levels throughout.
Adapting the scale for South African schools
The scale was designed in higher education settings with reliable connectivity. South Africa has a stark digital divide: some schools have world-class facilities, while many have no consistent internet. The adaptation matters more than the framework.
For schools with limited resources, No AI and the planning level are the practical core. Occasional, structured AI sessions in a computer lab work better than any expectation of constant access, and every task where AI is allowed should still be possible with printed materials or shared devices, so access does not decide the mark.
Better resourced schools can use all five levels and test approaches that others can learn from later.
One task at three levels
Take a Grade 10 essay on water scarcity in your community. At No AI, learners write it in class from notes they made earlier. At the planning level, they may use AI to generate questions and an outline, and hand in their prompts with the essay. At the collaboration level, they may ask AI for feedback on a draft, then submit the draft, the feedback and a short note on what they accepted, rejected and why. The topic is the same, but each version assesses something different, and learners know in advance which one they are doing.
How it fits South African assessment rules
School-based assessment is governed by the National Protocol for Assessment Grades R to 12. The AIAS does not replace it. It sits inside it, as a way for teachers to state the AI conditions for each internal task alongside the existing evidence and moderation requirements.
Policy is moving. In May 2026 the Minister of Basic Education, Siviwe Gwarube, wrote that the department is reviewing its digital education policy framework and developing practical guidance for schools. The Western Cape Education Department published an AI position paper for its schools in August 2026. Until national guidance arrives, a school-level AIAS policy gives teachers and learners clarity now. If your school does not yet have an AI policy at all, start with governance, not technology.
Common concerns
When I discuss the AIAS with South African educators, the same four concerns come up:
- Equitable access. Not all learners have AI tools at home. Scheduled computer lab time, or partnerships with community centres, can narrow the gap.
- Data costs. Data remains expensive. Education authorities could negotiate zero-rated access to educational AI platforms, as many did during pandemic remote learning.
- Language. South Africa has twelve official languages, including South African Sign Language since 2023, yet most AI tools work best in English. Teachers should be open about that limit while exploring how AI can still support multilingual learners.
- Teacher training. Many teachers have had no formal AI training. Practical professional development on using the scale is worth more than general awareness sessions. My guide to AI professional development for teachers sets out what works.
Where to start
Do not wait for perfect conditions. Pick one subject, agree the level for each of its tasks this term, tell learners before each task, and collect feedback from teachers and learners before extending it to other subjects. That builds confidence and gives you local evidence of what works.
Assessment is one part of a wider picture. To see where your school stands on leadership, policy, staff capacity and curriculum as well, use the AI readiness framework for schools.
I run sessions on AI and assessment for school leaders and education departments. If that would help your team, see speaking or get in touch.
More on AI and assessment
- Designing assessments with AI, my work with assessment professionals
- Teaching assessment professionals to work with AI
- Designing formative assessments with spreadsheets and AI
- Formative assessment: the core principles
Discover more from Niall McNulty
Subscribe to get the latest posts sent to your email.
2 thoughts on “Adapting the AI Assessment Scale for South African Schools”