Writing workshops at a South African university: what we're seeing with student AI use-and what actually helps
AI can draft fast, tidy sentences. That's the appeal. But in our writing workshops at a South African university, we're seeing a pattern: over-reliance on AI strips students of agency and voice.
For African universities, this matters. Student voice fuels social change and decolonisation. If AI becomes the writer, students stop thinking, stop questioning, and stop sounding like themselves.
What worries us most
AI doesn't just tidy prose; it can quietly replace the student's mind on the page. That's the risk. The minute a tool becomes the default author, curiosity and ownership fade.
When that happens, original thought is replaced by generic answers. The work might look polished, but it's hollow.
How we spot AI-written text
- Flawless grammar paired with clichés. The prose is smooth, but the ideas feel flat.
- Generic, shallow reasoning. Statements sound correct, but don't push the argument forward.
- No local grounding. Creative pieces drift to nameless settings with characters like "Stacey" or "Rick," instead of real South African places and names.
- Missed context. A Geography essay on sustainability discussed global themes-human rights, inequality, environmental justice-yet never mentioned South Africa, despite the brief asking for it.
- Students can't explain their own text. In consultations, they ask what their "essay" means-because they didn't write it.
- Detection tools are useful as a first flag, not a verdict. They help us know where to look closer.
Interventions that actually help
- Make students interrogate the text: Does this make sense? Is it true? Does it answer the exact question asked?
- Bring back pens and paper in controlled settings. It restores focus and shows who can think on their feet.
- Rebuild reading as a core skill. If you can't read critically, you can't judge what an AI produces.
- Teach simple, repeatable frameworks. For reading, the SQ3R method works well: survey, question, read, recite, review. Here's a solid overview.
- Teach paragraph structure with PIE: Point, Illustration, Explanation. It forces clarity and logic.
- Anchor tasks in indigenous knowledge and African contexts. AI tends to default to western and northern views; your brief must counter that.
- Model the behavior. If lecturers over-use AI for feedback or content, students will follow.
What we changed in our workshops
We've rebuilt sessions to focus more on reading for understanding. Strategies like SQ3R and PIE are taught and practiced, not just mentioned.
We spend more time asking students to explain what they wrote and why it answers the brief. If they can't explain it simply, they didn't own the process.
One-on-one consultations remain key. They help students tighten arguments, use sources properly, and keep their voice intact.
Practical guidance you can use right now
- Use AI for brainstorming or outlining, then switch it off. Write the first pass yourself.
- Localize every assignment. Require examples, sources, and cases from South Africa or the specific community you serve.
- Interrogate every paragraph with PIE. If you can't spot the Point, Illustration, and Explanation in 20 seconds, rewrite.
- Force reflection. Ask students to summarize their main argument in two sentences-no notes allowed.
- Treat AI detection as a starting point. The real test is oral explanation and source cross-checks.
AI can be a useful assistant, but the student must stay the author. That's the line. Keep the pen in their hand, and make the process-reading, thinking, drafting-non-negotiable.
If you want to sharpen how you brief AI without losing your voice, explore practical prompt techniques here: Prompt Engineering resources.
Your membership also unlocks:
