Show what matters now. Reveal the rest on demand.
Progressive disclosure defers advanced or rarely used options to a secondary layer, keeping the first screen focused on what most users need most of the time. It is the working compromise between power and simplicity.
The technique was studied at IBM in the early 1980s, where Jack Carroll’s training-wheels interface cut novice errors by hiding advanced functions until users were ready for them.
Jakob Nielsen named and popularized the pattern for interface design, calling progressive disclosure one of the best ways to satisfy both novice and expert users with a single interface.
Show users only what they need for the current task, and provide a clearly labeled path to the rest. Primary options stay immediate; advanced, rare, or dangerous options sit one deliberate step away — present, but not in the way.
The core move: a small primary set shown immediately, and a labeled door to everything else. Novices never meet the complexity; experts reach it in one click. Both groups use the same product.
The pattern lives or dies on what goes in layer one. Usage data decides: the handful of options that cover the great majority of tasks stay visible. Guess wrong and you’ve buried the product’s core.
The disclosure controls are a small family: the accordion, the “Show advanced” link, the overflow menu, the details drawer. Each needs an honest label — users must predict what’s behind the door before opening it.
A multi-step wizard discloses progressively across time instead of space: each step shows one decision’s worth of interface. The full form would terrify; the sequence feels manageable.
The pattern’s failure mode is over-hiding: when a frequently needed option lands in layer two, every user pays a click forever, and support tickets read “where did X go?” Frequency, not complexity, decides placement.
Tooltips, popovers, and expandable rows disclose depth in place — the summary stays scannable while the detail waits underneath. Shneiderman’s mantra applies: overview first, details on demand.
Chat answers and agent reasoning are disclosure problems: the short answer first, the working shown to those who ask.
A good model response is progressive disclosure in prose: the direct answer up top, the derivation expandable below. Dumping the full reasoning on everyone is the wall-of-form mistake, rewritten in text.
Agent traces, tool calls, and chain-of-thought belong behind a disclosure: visible enough to audit, folded enough to stay out of the way. Trust needs the door to exist; flow needs it closed by default.