AI Dictionary › Prompting
Prompt di Verifica
Verification prompting is a technique in which, after obtaining an initial response from the model, a subsequent prompt is formulated dedicated exclusively to checking its correctness, internal consistency or factual accuracy before accepting it as final. Verification becomes a step separate from generation.
The typical mechanism has the model first generate a response, then receive a prompt asking it to critically reread it, list any doubtful claims, check them one by one and correct them if necessary. Separating generation from verification reduces the risk that the model, still "anchored" to the reasoning it just produced, fails to notice its own errors.
This technique is useful whenever factual accuracy matters more than speed: checking numerical data in an analysis, verifying regulatory references in a legal text, or reviewing generated code to spot logical bugs. It also underlies workflows where a second prompt, or a second model, acts as a checker of the first response.
The term simply describes the application to prompting of the independent-verification principle, an idea long present in fields like quality control and adopted in prompting as structured techniques emerged to have the model check its own claims.
Grace gives a higher score when, after the first answer, you add an explicit verification step instead of accepting the output exactly as it came out.
From our network
AGORÀ Intelligence: Enterprise AI Governance Platform
Govern AI at scale: policies, adoption and measurable results on your data. Built for boards and C-suite.
Visit agora-intelligence.com →From the Agora Intelligence blog
📱 Download the Android app (beta) iOS coming soon
Say what you mean. Get what you need.
Grace Certified, the AI coach that trains and certifies your prompt engineering, by Agora Intelligence.