AI will look you dead in the eye and tell you a banana is a vegetable.
It just makes stuff up. They call it a “hallucination,” and it’s one of the biggest problem with AI today. I see a lot of businesses in Zimbabwe haven’t understood this yet.
But here’s the thing… the less frequently it happens, the more dangerous it becomes.
You use an AI tool, it gives you a few great answers. You start to trust it. You get a little lazy. You stop reading all the way through and start just skimming. You give it the benefit of the doubt. Next thing you know, you’re copy-pasting directly into your monthly report, into your company profile, or even worse… into a client proposal.
And that can be funny, but it can also completely undermine your credibility. Or even get you fired.
So, how do you fix this?
There’s an easy solution. A simple two-step process that I use with every client.
Step one: you have to let your AI know you’re evaluating its answers.
You basically tell it, “I will be checking this for accuracy.” This forces the AI to be more careful. It knows it has to work harder.
Step two: make it double-check its own work.
Before you even look at the answer, you can ask it to check its response for logic, for consistency, and for any potential biases.
I’ve included below, two short prompts for you. They’re tried and tested.
You can just copy-paste them right into your conversations with any AI. Especially when accuracy matters. When the truth matters.
These little additions will just make your answers better. More reliable.
Feel free to use them, adapt them, and make them your own.
Here are some text strings you can add to your prompts, especially when accuracy, safety, and consistency are important.
1. For Accuracy & Fact-Checking
Use this when you are generating content that includes facts, data, or historical information.
> Before finalizing the response, conduct a rigorous accuracy check. Scrutinize all factual statements, statistics, and specific data points. Cross-reference the information against reliable sources to ensure its validity. If any information is speculative or cannot be definitively verified, explicitly state this. The final output must be factually sound.
2. For Legal & Policy Compliance
Use this when creating content for regulated industries (like finance, healthcare, legal) or when you need to adhere to specific company policies.
> Perform a compliance review of the output. The content must adhere strictly to the following guidelines and restrictions: [INSERT YOUR SPECIFIC GUIDELINES HERE – e.g., “Do not make financial guarantees,” “Avoid giving medical advice,” “Ensure all marketing claims are substantiated.”]. Flag and rewrite any part of the response that could violate these rules.
3. For Brand Consistency
Use this to ensure the AI’s output matches your company’s or personal brand’s tone of voice and style.
> Analyze the output for brand consistency. The tone of voice must be [INSERT TONE – e.g., “professional yet approachable,” “energetic and witty,” “formal and academic”]. The language must align with our brand style, which includes [INSERT KEY RULES – e.g., “using ‘team members’ instead of ’employees’,” “avoiding technical jargon,” “writing in the active voice”]. The final text should feel as if it were written by a core member of our team.
4. For Bias & Inclusivity
Use this to ensure your content is fair, balanced, and appropriate for a diverse audience.
> Conduct a bias and inclusivity check on the response. Review the text for any stereotypes, generalizations, or assumptions related to age, gender, ethnicity, nationality, disability, or other personal attributes. Replace any non-inclusive or loaded language with neutral and respectful alternatives. The goal is to ensure the content is welcoming and fair to all readers.
Now, using these prompts doesn’t mean your AI assistance won’t hallucinate or “lie to you” – but it does significant improve your chances. Still do take a moment to double check.
And as always,
stay smart!
Max Soutter 
Say something!