A team has shipped an AI product that generates marketing copy. Users frequently get unexpected output (wrong tone, off-brand claims, occasional hallucinations) and report frustration. They're considering UX changes to improve trust without changing the underlying model.
Which UX patterns most directly address user trust in AI products?
Why did you pick that answer? Two or three sentences. The act of articulating it is what builds the judgment — not the click that follows.
AI products break user trust differently from non-AI products — outputs are non-deterministic, occasionally wrong, and users can't inspect why the output is what it is. The Set Expectations / Easy Recovery / User Control triad addresses each of these: expectations prevent surprise; recovery makes failures cheap; control gives users agency over a process they don't fully understand. Skipping these three produces products that feel magical when they work and infuriating when they don't — which is exactly the trust-erosion path the team is on.
Adding a chat layer doesn't fix bad output — it just adds another surface where users encounter it. Chat is a useful pattern for some AI products but not a fix for trust failures in others.
Hiding AI generation is both ethically problematic (users have a right to know) and counterproductive (when something feels off, users discover the AI nature and trust drops further). Transparency is a trust mechanism, not an obstacle.
Adding AI features to a product whose AI features are eroding trust compounds the problem. The fix is making the existing AI features trustworthy, not adding more.