What Is a Skeleton Screen?

A skeleton screen is a promise about layout: this is where things will be, and they are coming. It feels faster than a spinner because it gives the eye something to expect instead of something to watch.

What Is a Skeleton Screen? — Troiana insight cover

In short

A skeleton screen is a placeholder version of a page that shows the layout as grey blocks where text, images and controls will appear, displayed while the real content loads, so the transition from empty to full is gradual rather than a spinner giving way to a page. Studies of perceived performance find users judge skeleton screens as faster than spinners for the same load time, particularly for loads under a few seconds. They work when the final layout is predictable; they mislead when the content's shape is unknown, and they are worse than nothing for loads that take long enough to need a real progress indicator.

What it is

Instead of a blank area or a spinning icon while data loads, a skeleton screen renders the page's structure with placeholders: a grey rectangle where the image will be, grey bars where the lines of text will be, a rounded block for the button. Often the placeholders shimmer gently. When the content arrives, it replaces the placeholders in position, and the page appears to fill in rather than to appear.

Spinner
Loading…
Skeleton
A static illustration. The skeleton promises the layout, and reserves the space so nothing shifts when content arrives.

The pattern was described by Luke Wroblewski in 2013 and adopted by Facebook, LinkedIn, YouTube and most content-heavy apps since. It is now a standard component in design systems.

Why it feels faster

Perceived performance is about attention. A spinner draws the eye to the fact of waiting: the user watches it turn and counts the time. A skeleton draws the eye to the page: the user reads the layout, anticipates where the content will land, and the wait registers as part of the page arriving rather than as a separate delay. Studies and A/B tests, including published work from the companies that adopted the pattern, have found that users rate skeleton-loaded pages as faster than spinner-loaded ones with identical load times, and that the effect is strongest for loads of a few hundred milliseconds to a few seconds.

There is a second effect: skeletons reduce layout shift. Because the placeholders reserve the space, content does not jump into position and push other things around, which is both a measured Core Web Vitals improvement and a felt one.

Skeletons, spinners and progress bars

Skeleton: for content whose layout is known and whose load takes under a few seconds. Lists, feeds, cards, profile pages, dashboards.

Spinner: for short, indeterminate waits on actions rather than page loads, where there is no layout to preview: submitting a form, saving, a button's own pending state. Also acceptable for very brief page loads where a skeleton would flash.

Progress bar or percentage: for long or determinate waits, uploads, exports, processing, where the user needs to know how much remains and whether it is still moving. A skeleton that sits for fifteen seconds becomes a page that looks broken.

Nothing: for loads under about 100 milliseconds, where any indicator is a flash of noise. Delay showing a skeleton by a short threshold so it appears only if the wait is real.

Optimistic UI: for actions whose result is predictable, show the result immediately and reconcile later; this sidesteps loading states entirely for the common case.

Where skeletons mislead

Unknown layouts. A skeleton promises a shape. If the content turns out to be an error, an empty state, or a completely different layout, the promise was false and the switch is jarring. Skeletons suit content whose shape is predictable; for uncertain results, a neutral indicator is more honest.

Long waits. Past a few seconds a skeleton stops reassuring and starts looking like a frozen page. Escalate to a progress indicator or a message.

Mismatched skeletons. Placeholders that do not match the real layout, three lines where one appears, an image block where there is none, cause a jump on load that undoes the benefit.

Skeletons for everything. Every component with its own shimmering placeholder, appearing and disappearing at different times, is visually noisy. Skeleton the page's main content region, not every widget.

Hiding slowness. A skeleton is a presentation of a wait, not a reduction of it. If the load is slow, the fix is the load.

Building one that behaves

  • Match the final layout in dimensions and position; generate the skeleton from the real component with its content replaced, rather than drawing an approximation.
  • Reserve exact space for images and media so nothing shifts when they arrive.
  • Use a subtle, low-contrast placeholder in a neutral tone drawn from the palette, with a gentle shimmer or pulse, and disable the animation when the user prefers reduced motion.
  • Delay by 100–300 milliseconds so skeletons appear only for real waits and do not flash on fast loads.
  • Set a ceiling, after which the skeleton gives way to a progress indicator or a message.
  • Announce loading state to assistive technology with aria-busy on the region, and do not read the placeholders aloud.
  • Handle empty and error states explicitly, replacing the skeleton with the appropriate designed state rather than leaving grey bars.
  • Load in a sensible order: the region the user looks at first fills first.

The decision

Use a skeleton wherever a page with a predictable layout loads its content after the shell, keep it faithful to the real layout, and pair it with a fallback for long waits. It is one of the cheaper perceived-performance improvements available and one of the more often botched. If your product's loading states feel slower than they measure, book a call; the loading choreography is usually a day's work to fix.

Common questions

What is the difference between a skeleton screen and a spinner?

A spinner is an icon indicating that something is loading, with no information about what. A skeleton screen shows the layout of the coming content as placeholder blocks, so the page appears to fill in rather than to appear. Skeletons suit page and content loads with predictable layouts; spinners suit short waits on actions such as saving or submitting.

Do skeleton screens actually improve perceived performance?

Yes, for loads of a few hundred milliseconds to a few seconds. Studies and tests from companies that adopted them found users rate skeleton-loaded pages as faster than spinner-loaded pages with identical load times, because attention goes to the emerging layout rather than to a symbol of waiting. They also reduce layout shift by reserving space.

When should you not use a skeleton screen?

When the final layout is unpredictable, such as a result that may be an error or an empty state; when the wait is long enough to need a progress indicator; for very fast loads where any placeholder flashes; and for every small widget on a page, which becomes visually noisy. Use them for the main content region of pages with known layouts.

How long should a skeleton screen show before switching to something else?

Show it only after a short delay of 100 to 300 milliseconds so it does not flash on fast loads, and replace it after a few seconds with a progress indicator or a message, since a skeleton that persists looks like a frozen page. The exact ceiling depends on the product; around three to five seconds is common.

Are skeleton screens accessible?

They can be. Mark the loading region with aria-busy so screen readers know content is pending, prevent the placeholders from being read aloud, keep the shimmer subtle and disable it when the user prefers reduced motion, and ensure the loaded content is announced or focus is managed appropriately when it arrives.

Have something worth building right?