High CLS in field data but not in Lighthouse?

Lighthouse tests one page load without scrolling or tapping, while Chrome counts layout shifts for the whole visit. Most large shifts happen after load, on scroll or tap.

Arjen KarelArjen KarelUpdated November 22, 2026

Why is CLS high in field data when Lighthouse shows none?

In short

  • CLS can pass in Lighthouse and fail in Search Console because field data counts layout shifts for the whole visit, as long as the page stays open.
  • A Lighthouse navigation run loads the page once and never scrolls or taps.
  • To find the shift real visitors get, check when it happened, which element moved and how far down the page they had scrolled.

Search Console and the field section of PageSpeed Insights report Cumulative Layout Shift (CLS) from opted-in Chrome users, and a real-user monitoring (RUM) tool measures it in every Chromium browser. The Performance score in PageSpeed Insights comes from a single Lighthouse load. So both can be right about the same URL: a page that sits still while it paints can jump later, when a visitor scrolls into lazy-loaded content, taps a filter or gets a late banner (lab data vs field data).

Check first whether Search Console or PageSpeed Insights shows a group score instead of the URL you tested. Search Console groups similar URLs and PageSpeed Insights falls back to the whole origin when a URL has too few visits, so the shift can sit on another template (why PageSpeed Insights and your RUM differ). With no RUM on the site yet, record a scroll and a few taps in the Chrome DevTools Performance panel; how to identify and fix CLS has the steps.

Which layout shifts count after the page has loaded?

Every layout shift after the page loads counts toward CLS, shifts while scrolling included, unless it comes within 500 milliseconds of a tap, click or keypress. When a tap starts a network fetch and the new content pushes the page down 600 milliseconds later, Chrome counts the shift. A scroll never opens that window, and neither does pinch-zooming, dragging or hovering.

Chrome keeps watching until the tab is hidden or closed. CLS is the largest burst of shifts less than a second apart and up to five seconds long; good is 0.1 or less and poor is over 0.25 (how CLS is scored). Google judges that score at the 75th percentile of page views (what p75 means in Core Web Vitals).

Only Chromium browsers, such as Chrome and Edge, report layout shifts; Safari, Firefox and every browser on an iPhone or iPad report no CLS at all. A shift inside a cross-origin iframe counts in Google's Chrome UX Report (CrUX), but no script on your page can see it.

How often does the worst shift come after the load?

On phones, across 103 CoreDash projects in the 28 days to 21 November 2026, about three in four page views with a poor CLS (over 0.25, 224,544 in all) had their largest layout shift after the page had loaded. Of those with a recorded moment, 47% shifted after a scroll, 21% after a tap or keypress and 7% after the load with no input, against 24% during the load.

That split explains a common trap: a script-delay plugin fools Lighthouse, but your visitors still get the shift. It holds JavaScript back until the first scroll or tap, so banners, widgets and ads never load in a test that touches nothing. Real visitors do scroll and tap, and the delayed scripts then insert their markup where Chrome counts every shift.

How do you see when the shift happened for your own visitors?

On the CLS page in CoreDash, the "When it shifted" card puts every page view with a layout shift into one moment, taken from its largest shift. There are four moments plus "Not recorded": "During load, before any input", "After load, without input", "After a tap or keypress" on Mobile ("After a click or keypress" on Desktop and across all devices) and "After a scroll".

"After a tap or keypress" means the visitor's last input before the shift was a tap or keypress more than 500 milliseconds earlier; it doesn't prove the tap caused the shift. "After a scroll" works the same way. For the two raw signals on their own, the Pages table has "Load state" and "Last interaction" under More (CLS dimensions in CoreDash).

The longest bar holds the most page views over 0.10; the p75 column shows how bad each moment is.

The verdict above the rows names a rarer moment with a higher p75. "Per page" switches the Pages table below to "When it shifted"; a row's menu there has "Filter to this", which narrows the whole CLS page to that moment.

If a CoreDash issue card sent you here, this is the view behind it. "Layout shift after interaction" opens when CLS is over 0.10 both after a tap, keypress or scroll and across the page group. "Element that shifts" opens when one shifting element is over 0.10 across enough page views.

When Mobile is much higher than Desktop, or one Folder or URL stands out, that is a correlation: phones and desktops also differ in screen height, browser bars and ad placements. The split shows where to look.

Which element moved, and what did the visitor tap first?

The Elements card on CoreDash's CLS page names the element that moved in the largest shift of each page view, with the most page views over 0.10 first. Its When column shows the usual moment for that selector. When that is "After a tap" ("After a click" on Desktop), hover the cell to see which control visitors tapped most often before the shift. The Pages table can also break down by "Tapped element" under More (Tapped element and CLS element).

On a tap row, hover the When cell to see which control was tapped first.

The element that moved is not always the cause. Chromium reports the visible elements that got pushed; the new node that pushed them never moved, so it isn't listed. When a late banner, ad slot or filter bar expands, the heading or grid right below it shows up in Elements. Look at the element right above that selector in the DOM, or at the control tapped before the shift. "No element" means the browser named no element for the largest shift, or what moved was not an element.

How far down the page did the layout shift happen?

The "Scroll depth" tab under More on CoreDash's CLS page shows where the bottom of the screen had reached at the largest layout shift, as a share of the page height in steps of 5% (100% when the whole page fits the screen). Across 97 CoreDash projects in the same 28 days, of the 98,765 poor shifts after a scroll on phones, one in three came when the visitor was halfway down the page or further.

Filtered to "After a scroll", the top rows are the depths to scroll to.

To find that shift on your own page, click "Per page" on the "When it shifted" card, open the menu on "After a scroll" and choose "Filter to this", then switch to "Scroll depth" (Scroll depth in CoreDash). Filter to one URL as well, open it on a phone-sized screen and scroll until the bottom of the screen reaches the depth with the most page views over 0.10. The lazy image, ad slot or collapsible section around that point is where to look.

What causes a shift after the load, and how do you fix it?

Match the moment that holds most of your field data to what the page inserts or resizes at that point.

During load, before any input

Shifts during load show up in field data when slow mobile connections expose unsized images or a late web font swap that a single test run got through cleanly. Turn on automatic image dimensions in your CMS or media plugin, and ask your developer to match the fallback font's metrics so text doesn't reflow when the web font arrives.

<img src="/hero.jpg" width="1200" height="675" alt="Product overview">

Responsive images and font metric overrides are in Cumulative Layout Shift on corewebvitals.io.

After load, without input

Shifts after the load without any input usually come from a cookie consent bar, a promo banner, a chat launcher or a late ad slot. In your consent, chat or banner app, switch a push-down bar to an overlay or a bottom bar. For an ad network or embed you can't edit, ask your ad team for the tallest creative that slot serves, turn off empty-slot collapsing and reserve that height on the wrapper.

.ad-slot { min-height: 250px; }

How to identify and fix CLS covers ad wrappers and embeds you don't control.

After a tap or keypress

Shifts after a tap or keypress usually come from a "Load more" button, a product filter, an accordion or a single-page route change whose update lands more than 500 milliseconds after the tap. Switch a dynamic product grid to paginated URLs or fixed-size skeleton cards in your theme or filter plugin, or ask the widget vendor for a fixed-height container. A loading animation doesn't hide a layout shift, it moves it: a small spinner right after the tap falls inside the 500-millisecond window, and Chrome still counts the shift when the results replace it. A placeholder that goes in on the tap, already as tall as the results, keeps the change inside the window.

.results-placeholder { min-height: 640px; }

How to identify and fix CLS shows how to test a slow tap locally.

After a scroll

Shifts after a scroll usually come when the visitor reaches a lazy-loaded image, iframe or mid-page ad with no reserved space, a section with content-visibility: auto and no placeholder height, or a sticky header that animates top or height instead of a CSS transform. Keep image dimensions on in your lazy-loading plugin, turn off header-shrinking effects in your theme and give in-article ad wrappers a fixed minimum height.

.sticky-header.is-hidden { transform: translateY(-100%); }

Pair content-visibility: auto with contain-intrinsic-size so unrendered sections keep their height. Sticky headers and transitions are in scroll-triggered animations that cause CLS.

Once a fix is live, check over the next days that the moment's bar in "When it shifted" shrinks and the selector drops down the Elements card. When one burst held several shifts less than a second apart, the next one takes the top spot until each is fixed (checking that a CLS fix worked).

If you own the site, start with the moment that holds your failing visits and the route under it. The developer uses Elements, "Tapped element" and "Scroll depth" to find the node right above the element that moved, and reserves its space before it paints or within 500 milliseconds of the tap. I treat a CLS of 0 as the goal on every site, even when 0.1 or less already passes: a shift inside the good range still moves the page under a reader's eyes.

For more, corewebvitals.io covers how Cumulative Layout Shift works and how to reproduce a shift in Chrome DevTools; our guides explain what p75 means and why CrUX and your RUM differ.

Sources: Optimize Cumulative Layout Shift (web.dev), Cumulative Layout Shift (web.dev), Layout Instability API (WICG), CrUX methodology (Chrome).

Track your first site free.No card needed. Setup takes about 5 minutes.

Start free trialBook a demo