In this article, we will highlight the top 10 common AI prompting mistakes, show bad prompt examples, and explain how the SAFE method helps prevent them.
1. Being Too Vague
- Bad Prompt Example: “Write a report.”
- Problem: No context, so the AI produces something generic and unhelpful.
- SAFE Fix: Set Context → “Write a 500-word report for SMEs on AI compliance, with examples.”
2. Forgetting the Audience
- Bad Prompt Example: “Summarise this policy.”
- Problem: AI doesn’t know who the summary is for—staff, customers, or executives?
- SAFE Fix: Set Context → “Summarise this policy in plain language for employees with no legal background.”
3. Asking for Too Much at Once
- Bad Prompt Example: “Write a blog, make it funny, persuasive, and highly technical.”
- Problem: Conflicting instructions confuse the AI.
- SAFE Fix: Ask Clearly → Break into smaller tasks: draft blog → refine tone → add humour where appropriate.
4. Ignoring Compliance Risks
- Bad Prompt Example: “Draft a press release about our new AI product—make bold claims.”
- Problem: Risk of exaggeration or regulatory breaches.
- SAFE Fix: Evaluate → “Draft a 300-word press release about our new AI product, using only verified information from this document.”
5. Copy-Pasting Sensitive Data
- Bad Prompt Example: “Analyse this file with customer names and financial data.”
- Problem: Potential data leakage or GDPR breach.
- SAFE Fix: Evaluate → Strip sensitive details before prompting, or ask AI to simulate analysis using placeholders.
6. Skipping Feedback Loops
- Bad Prompt Example: “Write a blog post on AI governance.” → Accepting the first draft.
- Problem: Missed opportunity to refine clarity and accuracy.
- SAFE Fix: Feedback Loops → “Good start. Can you shorten it, add a checklist, and keep the tone professional?”
7. Over-Relying on AI for Facts
- Bad Prompt Example: “Explain the latest EU AI Act requirements.”
- Problem: AI may hallucinate or provide outdated information.
- SAFE Fix: Evaluate → Cross-check outputs against official sources before use.
8. Not Defining Format
- Bad Prompt Example: “Create a training guide.”
- Problem: AI could produce text when you needed bullet points, slides, or a Q&A format.
- SAFE Fix: Ask Clearly → “Create a one-page training guide in bullet points, with a quiz at the end.”
9. Asking AI to “Be Creative” Without Boundaries
- Bad Prompt Example: “Write a funny sales pitch.”
- Problem: Humour can miss the mark, sound unprofessional, or even offend.
- SAFE Fix: Set Context + Ask Clearly → “Write a light, professional sales pitch with one joke suitable for LinkedIn.”
10. Forgetting to Review Outputs
- Bad Prompt Example: Using AI-generated text without editing.
- Problem: Errors, bias, or compliance risks slip through.
- SAFE Fix: Evaluate → Always review and adapt before sharing externally.
Conclusion
Most AI prompting mistakes stem from unclear instructions, missing context, or skipping the review step. These errors can lead to wasted effort, or worse, compliance risks.
By applying the SAFE method (Set Context, Ask Clearly, Feedback Loops, Evaluate), business teams can avoid the pitfalls of bad AI prompts and turn AI into a reliable, responsible tool for daily work.