Trust: Prove sources, show reasoning, and make consent inspectable before users act.
Essential
More in Trust
- Data Ownership & ControlUser control over AI data usage
- Source BrowserInspect retrieved sources and context beside the answer
- Confidence IndicatorsScores, meters, and badges for how sure the model is
- Progress StepsCollapsible thinking and tool traces
- Verification Next StepsConcrete actions to validate uncertain output
- Knowledge GraphExplore entities and links behind retrieved knowledge
- Privacy FiltersMask sensitive data in prompts, context, or outputs
- Bias DetectionFlag potentially biased outputs
- Audit TrailComplete log of AI decisions and data usage
- Transparency ReportPeriodic reports on AI behavior/accuracy
- Authentication ChainsLegible identity trails across agent actions
- Responsibility AttributionTrace which agent or human caused each action
- Agent IdentityStable name, version, and capabilities per agent
Frequently asked questions
Which trust patterns should ship first?
Lead with provenance (citations or source browser), explicit uncertainty, and a human escalation path. Add cost, autonomy, and audit signals when actions have side effects, spend, or compliance requirements.
When are trust patterns required versus nice-to-have?
Treat them as required for search, finance, health, legal, and enterprise knowledge products. Anywhere a wrong answer has external consequences. Consumer creative tools may start lighter but still benefit from confidence cues on high-stakes outputs.
How do citations differ from chain-of-thought UI?
Citations tie claims to external sources users can verify. Chain-of-thought shows reasoning steps the model took. They solve different doubts (source vs logic). Use both when answers are long or disputed.
What anti-patterns hurt trust in AI interfaces?
Fake certainty, buried sources, inconsistent confidence, and silent data use. Avoid decorative “trust” badges without actionable provenance or controls.
How do trust patterns connect to agentic products?
Agents amplify risk because they act, not only reply. Pair trust patterns with autonomy budgets, approvals, and audit trails from the Agentic UX framework so users can see and revoke what ran.
Do pattern pages include demos for trust flows?
Many trust patterns include interactive demos plus screenshots from products like Perplexity and Google AI Overviews. Open the essential patterns above for the highest-traffic conventions.