AI failure states UX
AI failure states UX covers how products behave when models refuse, hallucinate, timeout, or lose tool access: honest disclosure, recovery paths, and human escalation instead of silent failure. Plan for wrong answers before you ship the happy path.
Start here
Patterns to ship first
Guides & frameworks
Production failure UX
Frequently asked questions
What are AI failure states in UX?
Failure states are UI for when AI cannot complete a task: refusals, tool errors, timeouts, low confidence, or policy blocks, distinct from generic network errors.
Should AI failures look like app errors?
Partially. Use familiar error patterns, but add AI-specific context: what was attempted, what data was missing, and what the user can do next (retry, edit prompt, escalate).
What is failure disclosure?
Failure disclosure is honestly stating limits and misses in-line. "I cannot book this" beats a fake success or vague "something went wrong."
How do recovery strategies differ from retries?
Retries repeat the same call. Recovery strategies change approach: fallback models, alternate tools, simplified output, or human handoff.
Which patterns reduce harm from wrong answers?
Failure disclosure, human checkpoints before irreversible actions, regen/alternatives, and verification next steps, not only accuracy improvements behind the scenes.