Telling people what the thing can and cannot do, before they find out the hard way. Overclaiming produces a spike of use followed by abandonment when the first confident wrong answer appears. Being specific about limits earns more durable trust than being impressive.
Technical view
Mechanisms include onboarding that demonstrates both strong and weak cases, in-context capability statements, clear scoping of what data the system can see, and consistent labelling of generated content. Evidence from automation research indicates that accurate prior expectations improve appropriate reliance and reduce both disuse after a single failure and over-reliance after early success. Overclaiming is also increasingly a consumer protection exposure.
Related
Part of AI Terminology Index, section: Product and Adoption.