Academy
1. What Is the TrustLayer Badge?
Understand that every AI answer receives an independent trust score out of 100 with a color code, and that it is a reading aid rather than a certificate of truth.
Every time an AI generates an answer on OUPI, TrustLayer independently evaluates it and displays a badge. This badge shows a score out of 100 and a color:
• Green – all checks pass. • Orange – at least one check raised a warning. • Red – a serious problem was detected.
The badge is a reading aid, not a certificate of truth. It tells you how consistent the answer is with the sources, your request, and the tools used — but a statement that is consistent with those inputs yet false in the real world can still receive a green score. Your own judgment remains essential for decisions and commitments.
Behind the single score, TrustLayer runs five independent controls:
- Source anchoring – Are the answer's claims actually found in the documents, knowledge bases, or web results it relies on?
- Prompt fidelity – Does the answer cover what you asked and respect your constraints?
- Action claims – When the AI says it performed an action (sent, saved, searched), did that action really happen?
- Code guard – Does any generated code contain dangerous patterns?
- Repetition – Is the model recycling the same content instead of producing a fresh response?
Each control contributes to the overall score and color. You can click the badge to see the result of every control individually.
Not every reply gets a score. Short conversational answers — like "sure, which one?" — contain no verifiable claim. Scoring them would dilute the badge's meaning, so TrustLayer marks them as not applicable instead. This way, when you do see a score, you know the system had something substantive to evaluate.
An orange or red badge is an invitation to check, not a reason to panic. Orange means one control flagged something worth reviewing (e.g., a claim not found in sources). Red signals a serious issue such as an action claimed but never performed. Ask the AI to cite its sources, rephrase your request, or attach a document to ground the answer. Never forward a red-badge answer without verifying it first.
Want a higher trust score? Ground your conversation. When you attach documents or connect a knowledge base, TrustLayer can verify claims against those sources (source anchoring). Ungrounded, open-ended questions give the system less to anchor against, so scores may be lower or controls less decisive.
Open a chat on OUPI and ask a factual question — for example, ask the AI to summarize a document you upload. Once the answer appears, look for the TrustLayer badge. Click it to expand the details: review each of the five controls and note which ones passed and which flagged warnings. Try rephrasing your question or adding a source to see how the score changes.
Every AI answer on OUPI receives a TrustLayer badge: a score out of 100 with a green, orange, or red color. Five controls — source anchoring, prompt fidelity, action claims, code guard, and repetition — feed that score. Short conversational replies are marked not applicable. Green does not mean "true in the world"; it means consistent with visible sources, your request, and the tools used. Treat orange and red as prompts to verify, and ground your questions with documents to help TrustLayer do its best work.