For years, people were told to look for spelling mistakes, strange grammar and awkward formatting when checking phishing messages.
That advice was never perfect, and AI is making it even less reliable.
In a July 2026 scam alert, SARS specifically noted that criminals were using AI to generate professional-looking email templates.
The important point is not that AI magically makes a scam undetectable. It makes it easier to produce convincing text quickly and repeatedly.
That means users should move away from judging a message only by how polished it looks.
Instead, check the request itself.
Were you expecting the message? Does it ask you to click a link and sign in? Is it creating urgency? Is it asking for money or sensitive information? Can the request be verified through a separate, trusted channel?
A convincing message can still be fraudulent.
This is especially important when the message appears to come from a bank, government department, employer, delivery company or another organisation you already trust.
The safest habit is independent verification. Open the organisation's official website yourself, use a known telephone number or use an existing app rather than following an unexpected link.
AI can improve a scammer's writing. It does not change the fundamental defence: verify before you trust.