Academy
3. Reading the Details Behind a Score
Learn how to click the badge to see each control's result and, for source anchoring, which specific claims were or were not found in the sources.
When you see a TrustLayer badge on an AI answer, it summarizes the outcome of five independent controls into a single score out of 100 and a color — green, orange, or red. But the real value comes from looking behind that summary. Click the badge and it opens a detailed panel listing each control individually: Source Anchoring, Prompt Fidelity, Action Claims, Code Guard, and Repetition. Each control shows its own result, so you can pinpoint exactly where the answer is strong and where it needs attention. This is especially useful when the badge is orange: instead of guessing what went wrong, you see which specific control raised the warning.
The Source Anchoring control deserves special attention because it gives you the most granular feedback. When an answer relies on documents, knowledge bases, or web results, this control checks whether each claim in the answer can actually be found in those sources. In the detail panel, you will see a breakdown: claims that were successfully matched to a source, and claims that were not. A claim marked as 'not found' doesn't necessarily mean it's false — it means TrustLayer could not locate supporting evidence in the materials available. This distinction matters: it tells you exactly where to focus your own verification effort rather than re-reading the entire answer.
Understanding the other controls in the detail panel helps you act faster. Prompt Fidelity tells you whether the answer actually covered what you asked and respected your constraints — useful when you gave a complex, multi-part prompt. Action Claims flags cases where the AI said it performed an action (sent, saved, searched) but the action didn't actually happen. Code Guard catches dangerous patterns in generated code. Repetition detects recycled content. Each control's result is independent, so a green Source Anchoring score can coexist with an orange Prompt Fidelity warning. Always read the full list before deciding your next step.
When Source Anchoring flags unmatched claims, try grounding your conversation: attach a document or connect a knowledge base before asking again. Answers built on your own sources give TrustLayer more material to verify against, which raises the anchoring score and makes the detail panel far more informative.
Remember: a green badge means the answer is consistent with the sources, your request, and the tools used — not that it is true in the absolute sense. Always apply your own judgment on important decisions, even when every control in the detail panel shows a pass.
Open a chat, ask a question that references a document or knowledge base, and wait for the answer. Click the TrustLayer badge next to it. In the detail panel, find the Source Anchoring section and identify which claims were matched and which were not. Try rephrasing or grounding the unmatched claims to see the score change.
Clicking the TrustLayer badge reveals the result of each of the five controls, not just the overall score. For Source Anchoring, you see exactly which claims were or were not found in the available sources. The other controls — Prompt Fidelity, Action Claims, Code Guard, and Repetition — each report independently, letting you pinpoint issues fast. Use this detail to decide where to verify, rephrase, or ground your request with documents. And always remember: green means consistent with what TrustLayer can see, not guaranteed true.