When is a tooltip the right place for help vs a sign the UI is unclear?
On this page
A tooltip is the right place for help when it adds supplementary detail to a control the user already understands, and it is a red flag when it is required to explain what the control even does. The test is simple: does the interface work without it. If you removed every tooltip and users could still operate the screen, the tooltips are doing their proper job, enriching. If removing them would leave people unable to tell what a control is for, the tooltips are not helping; they are propping up a design that failed to communicate on its own. Tooltips should enrich, not rescue.
The reason this distinction matters is that a tooltip is the weakest channel of help in the interface. It is hidden until hovered, unreachable by many touch users, invisible to anyone scanning the screen, and easy to miss even when present. Putting essential meaning there means hiding the one thing the user most needs behind the one place they are least likely to look. When the meaning is essential, it belongs in the visible interface itself, as a clear label, a recognizable icon, a sensible grouping. The tooltip is for the second layer, the detail a confident user might want and the rest can safely never see.
A designer recognizes the rescue tooltip in the icon-only toolbar where every button needs a hover to reveal what it does. If a row of glyphs is incomprehensible until you hover each one, the tooltips are not enriching anything; they are the only thing keeping the toolbar usable, which means the icons themselves have failed. Contrast that with a clearly labeled “export” button whose tooltip adds “exports the current view as CSV,” or a form field labeled “API rate limit” with a tooltip clarifying the unit. In both, the user already knows what the control is from the visible interface, and the tooltip merely deepens it. Remove those tooltips and nothing breaks; remove the toolbar’s and it collapses.
One real exception is that supplementary help is genuinely valuable and should not be confused with the patch. Some information legitimately belongs in a tooltip precisely because it is secondary: a definition of a term the control already names, the format an already-labeled field expects, the consequence of an already-understood action. The danger is only when the tooltip carries the primary meaning. There is also a related but separate call about whether something should be a permanent inline label rather than a hover at all, which is its own decision; the test here is narrower and prior to that, just whether the control would be understood without the tooltip’s help.
When you find yourself reaching for a tooltip, ask whether the control works without it. If the answer is no, do not write the tooltip; fix the control so its purpose is clear from its label, icon, or placement, and let the tooltip return only as enrichment once the control stands on its own. Caption the confusion and it stays confusing for everyone who never hovers; repair the control and the tooltip becomes the bonus it was meant to be.