AI Output Verification

Apply the VERIFY framework to validate every AI output before use in production.

Category: AI Literacy | Type: Skills

Skills: Verification, Quality Assurance, Trust

Techniques: Structured Output, Chain-of-Thought, Self-Verification

Prompt

Every AI output should pass through a verification framework before use. The VERIFY method: 1. V — Validate Format: does the output match the requested structure? Parse it programmatically if possible. 2. E — Evidence Check: for every factual claim, can you trace it to a reliable source? No source, no trust. 3. R — Reasoning Audit: does the logic chain hold? Ask the model to explain its reasoning step by step and look for gaps. 4. I — Internal Consistency: do different parts of the output contradict each other? Check conclusions against stated premises. 5. F — Freshness Check: is this information potentially outdated? Check the model's knowledge cutoff against your needs. 6. Y — You Decide: after all checks, apply human judgment. Does this pass the smell test? Would you stake your reputation on it? Automation ladder: for low-stakes tasks, spot-check 10%. For medium-stakes, verify every output. For high-stakes, require two independent verification methods. Build verification into [your workflow], not as an afterthought.

Browse all prompts at Ask Wisely