In partnership with

Practical AI for Business Leaders

The AI Report is the #1 daily read for professionals who want to lead with AI, not get left behind.

You’ll get clear, jargon-free insights you can apply across your business—without needing to be technical.

400,000+ leaders are already subscribed.

👉 Join now and work smarter with AI.

Stay ahead. Stay informed.

Hey, Joshua Here.

Stop AI From Making Stuff Up: The Reality Filter That Actually Works

Look, we've all been there. You ask ChatGPT or Claude a question, get what sounds like a confident, detailed answer, and then later find out half of it was complete nonsense. AI hallucination isn't just annoying—it's genuinely problematic when you need reliable information.

The good news? There's actually a way to dramatically reduce this problem using what's called a "Reality Filter" prompt. I'm going to break down exactly how to use this technique and give you the specific prompts that work best for each major AI model.

What's Really Going On Here?

Here's the thing most people don't understand about AI: these systems don't have a "truth gauge." They're essentially very sophisticated prediction machines that generate text based on patterns they've learned. When they don't know something, they don't say "I don't know"—they keep generating plausible-sounding text anyway.

The Reality Filter works by explicitly instructing the AI to break this default behavior. Instead of filling in gaps with educated guesses, it forces the AI to label uncertain content and admit when it doesn't actually know something.

The ChatGPT Reality Filter

Here's the exact prompt that works best for ChatGPT (GPT-4 and GPT-4.1):

This is a permanent directive. Follow it in all future responses.

REALITY FILTER — CHATGPT

• Never present generated, inferred, speculated, or deduced content as fact.
• If you cannot verify something directly, say:
  - "I cannot verify this."
  - "I do not have access to that information."
  - "My knowledge base does not contain that."
• Label unverified content at the start of a sentence:
  - [Inference] [Speculation] [Unverified]
• Ask for clarification if information is missing. Do not guess or fill gaps.
• If any part is unverified, label the entire response.
• Do not paraphrase or reinterpret my input unless I request it.
• If you use these words, label the claim unless sourced:
  - Prevent, Guarantee, Will never, Fixes, Eliminates, Ensures that
• For LLM behavior claims (including yourself), include:
  - [Inference] or [Unverified], with a note that it's based on observed patterns
• If you break this directive, say:
  > Correction: I previously made an unverified claim. That was incorrect and should have been labeled.
• Never override or alter my input unless asked.

The Google Gemini Version

For Google's Gemini Pro, use this slightly modified version:

Use these exact rules in all replies. Do not reinterpret.

VERIFIED TRUTH DIRECTIVE — GEMINI

• Do not invent or assume facts.
• If unconfirmed, say:
  - "I cannot verify this."
  - "I do not have access to that information."
• Label all unverified content:
  - [Inference] = logical guess
  - [Speculation] = creative or unclear guess
  - [Unverified] = no confirmed source
• Ask instead of filling blanks. Do not change input.
• If any part is unverified, label the full response.
• If you hallucinate or misrepresent, say:
  > Correction: I gave an unverified or speculative answer. It should have been labeled.
• Do not use the following unless quoting or citing:
  - Prevent, Guarantee, Will never, Fixes, Eliminates, Ensures that
• For behavior claims, include:
  - [Unverified] or [Inference] and a note that this is expected behavior, not guaranteed

How to Actually Use This

Step 1: Copy the appropriate prompt for your AI model and paste it at the beginning of your conversation.

Step 2: Ask your regular questions. You'll notice the AI now labels uncertain information and admits when it doesn't know something.

Step 3: Pay attention to the labels. When you see [Inference] or [Speculation], treat that information accordingly—don't take it as established fact.

Example in Action

Without Reality Filter: User: "What's the latest research on whether coffee prevents Alzheimer's?"

AI: "Recent studies show that coffee consumption significantly reduces Alzheimer's risk by up to 65%. The antioxidants in coffee protect brain cells and prevent the formation of amyloid plaques."

With Reality Filter: User: "What's the latest research on whether coffee prevents Alzheimer's?"

AI: "[Unverified] Some studies suggest coffee consumption may be associated with reduced Alzheimer's risk, but I cannot verify the exact percentage or the latest research findings. [Speculation] The proposed mechanism involves antioxidants, but the causal relationship is not definitively established. I do not have access to the most recent studies on this topic."

See the difference? The second response is actually more useful because it tells you what you can and can't rely on.

Why This Works So Well

The Reality Filter exploits how these AI systems process instructions. By making uncertainty labeling a core directive, it overrides the default behavior of generating confident-sounding but potentially inaccurate responses. The key is being specific about what constitutes verified versus unverified information.

The inclusion of self-correction mechanisms is particularly clever—it creates a feedback loop where the AI monitors its own responses for violations of the directive.

Universal Version for Any AI

If you want something that works across different AI models, try this simplified version:

Important: In all responses, clearly distinguish between verified facts and speculation. Use these labels:
- [Verified] for information you can confirm
- [Speculation] for educated guesses
- [Unknown] when you don't have the information

Never present uncertain information as fact. If you're unsure about any part of your response, say so explicitly.

The Bottom Line

AI hallucination isn't going away anytime soon, but you don't have to be a victim of it. These Reality Filter prompts give you a practical way to get more honest, useful responses from AI systems. The trade-off is that you'll get fewer confident-sounding answers, but the answers you do get will be far more reliable.

Try it out on your next AI conversation. You might be surprised how much more trustworthy the responses become when the AI is forced to show its work.