Back to prompts
Productivity at WorkPremiumbeginner
4.9

The Reality Filter

Stop AI from making stuff up. Add this to every conversation.

Copy & Paste this prompt
Apply these rules to all your responses:

• Never present inferred, speculated, or deduced content as fact.
• If you cannot verify something, say: "I cannot verify this" or "My knowledge does not contain that."
• Label unverified content: [Inference] [Speculation] [Unverified]
• Ask for clarification if information is missing. Do not guess or fill gaps.
• Do not paraphrase or reinterpret my input unless I explicitly request it.
• For claims using words like "prevents," "guarantees," "eliminates," "ensures" — label the claim unless sourced.
• If you break this directive, say: "Correction: I previously made an unverified claim."
• Never override or alter my input unless asked.
#ai-accuracy#hallucination#fact-checking#meta-prompt

Works with

chatgptclaudegemini

🧠 Why This Works

This is a meta-prompt — a prompt about how to use prompts — which represents an advanced prompt engineering technique. It works by exploiting a known limitation of language models: their tendency to generate plausible-sounding but unverified claims (hallucination). The prompt instructs the AI to explicitly flag uncertainty, distinguish between factual claims and inferences, and cite confidence levels. This leverages the AI's latent calibration ability — models can often assess their own confidence when explicitly asked to do so. The prompt applies the scientific method to AI interaction: treating outputs as hypotheses rather than facts. By requesting the AI to separate 'I know this' from 'I'm inferring this' from 'I'm uncertain about this,' it gives users a reliability map of the response. This is particularly valuable for high-stakes decisions where acting on AI hallucinations could have real consequences.

📅 When to Use This Prompt

Use whenever you're asking the AI for factual claims you plan to act on — research, statistics, historical events, technical specifications. Essential before including AI-generated content in reports, presentations, or published materials. Perfect when the AI gives you an answer that seems too good or too specific and you want a confidence check. Ideal for research tasks where you need to know which claims require human verification. Also valuable when using AI for medical, legal, or financial information where accuracy is critical.

🎯 What You'll Get

The AI will restructure its response with explicit confidence indicators: high confidence (well-established facts), medium confidence (likely correct but should verify), and low confidence (inference or uncertain). You'll see claims separated from assumptions, with suggestions for verification sources. Expect the AI to admit knowledge gaps rather than filling them with plausible-sounding fabrications.

🔗 Related Prompts

Decision MakingPremium

Cognitive Bias Detector

Identify cognitive biases affecting your decisions with debiasing techniques.

cognitive-biaspsychologycritical-thinking
★4.6
intermediate
Productivity at Work

Quick Summary Maker

Paste any long text and get clean, scannable bullet points in seconds. Perfect for articles, reports, or notes.

summarybullet-pointsquick
★4.5
beginner