Tap delay

When the first iPhone put a desktop browser on a phone, it had to make pages built for a mouse usable at a quarter of the width, and its answer was double tap to zoom. That single decision created the delay. A tap arrives, and the browser cannot yet know what it is: on its own it is a click, but if a second tap lands within about three hundred milliseconds the pair means zoom instead. So the browser waited out the window before dispatching the click, and every button on the mobile web answered a third of a second late. That is long enough to feel broken rather than slow, which is why people tapped again, and why interfaces of the era were full of double-fired actions.

The workaround came first and the fix came later. Libraries such as FastClick listened for the raw touch events, synthesized a click immediately, and swallowed the real one when it finally arrived, which worked and brought problems of its own, including the ghost click: a click dispatched at the coordinates of a touch after the interface underneath had already moved on. Browsers then removed the delay properly, one condition at a time. A page declaring <meta name="viewport" content="width=device-width"> is already laid out for the screen, so double tap to zoom buys it nothing and Chrome stopped waiting on such pages in 2015; Safari followed on iOS 9.3. touch-action: manipulation says the same thing for a single element, keeping panning and pinching while giving up the double tap, and it remains the precise tool for a control on a page that is otherwise zoomable.

The delay is gone from any page that meets those conditions, so a script that fights it in 2026 is a script fighting a browser it will never meet. FastClick was retired for exactly this reason, and its ghost clicks and its interference with form controls went with it. The term survives because the symptom outlived the cause: an unresponsive tap is still the first thing people call the tap delay, and it is now almost always something else, a heavy event handler blocking the main thread, an animation that starts before the state changes, or a control with no pressed state at all to acknowledge the touch.

The lesson worth keeping is about zoom, not about milliseconds. Every fix listed above works by giving up a gesture, and the ones that give up too much are the ones that cause harm. user-scalable=no and a maximum scale of one also remove the delay, and they remove pinch to zoom along with it, which fails WCAG and takes away the accommodation people rely on most. Set the viewport width, reach for touch-action: manipulation where a control needs it, and leave zooming alone.