How to Stop Layout Shift on Ad Heavy Game Guide Pages
The complaint that finally made me fix this arrived in our wiki Discord as a seven second clip. A reader had opened our raid boss guide on an Android phone, started reading step two of the phase one rotation, and watched the paragraph drop...

The complaint that finally made me fix this arrived in our wiki Discord as a seven second clip. A reader had opened our raid boss guide on an Android phone, started reading step two of the phase one rotation, and watched the paragraph drop 250 pixels down the screen as an ad slot filled. He scrolled back up, found his place, and the related guides widget loaded underneath and pushed everything again. His message was four words long and none of them were polite.
Our Search Console report had been showing 68 percent of mobile URLs as needing improvement for Cumulative Layout Shift for about five months, and I had been ignoring it because Lighthouse gave the same pages a CLS of 0.00 on my laptop. Both readings were accurate. One measured a page nobody uses, loaded once, on a fast connection, without scrolling. The other measured what actually happens to readers.
What CLS is counting, and why your laptop lies about it
Cumulative Layout Shift scores how much visible content moves without the reader causing it. Each shift gets an impact fraction, roughly how much of the viewport moved, multiplied by a distance fraction, how far it travelled. The score you are judged on is the largest burst of shifts inside a five second window, and the threshold for good is 0.1. Above 0.25 counts as poor.
The gap between lab and field numbers comes from what gets measured. A Lighthouse run loads the page, waits, and stops. It does not scroll. Ads below the fold never request, so they never shift anything, so the score is clean. Field data from real Chrome users covers the whole visit, including the scroll down through four ad slots that each reserve nothing. Our page with a perfect lab CLS was sitting at 0.31 in the field, and that number is the one search uses. I wrote more about that mismatch in why page speed scores mislead community admins, and ad slots are the clearest example of it.
The practical consequence: you cannot audit this from a score. You have to watch the page move.
Finding every shift on a real guide page
Chrome DevTools has the tool for this and almost nobody uses it. Open the Rendering drawer, tick Layout Shift Regions, then reload the page and scroll slowly from top to bottom on a throttled connection. Every shift flashes a blue overlay exactly where it happened. On our boss guide there were nine. Three were ads, two were images without dimensions, one was the author avatar, one was the cookie banner, and two were the font swap retypesetting the first screen of text.
Then open the Performance panel, record the same scroll, and look at the Experience track. Each shift there is clickable and tells you the moving element and the score it contributed. That is how I learned that our worst single shift was not an ad at all. It was a 0.19 jump from the tier list table restyling itself once a web font finished loading.
Field data matters just as much. In Search Console, the Core Web Vitals report groups URLs rather than listing them, so pick the example URL from the failing group and run it through the Chrome User Experience report rather than trusting one phone. Guide pages and news pages on the same site usually fail for completely different reasons, and a template level fix applied to the wrong template wastes a week.
The fix everyone recommends, and why it backfires
Search for this problem and you will be told to put a fixed height on every ad container. Give the slot a 250 pixel min height, the ad arrives, nothing moves. It works, and on a guide site it costs you more than it saves.
Fill rates are the reason. A fan wiki running a small network does not sell every impression. Our unsold rate on mobile sat near 22 percent at the time, and on tablet it was worse. Reserving 250 pixels for every slot means that on roughly one page view in five the reader scrolls through a stack of empty boxes, three or four of them on a long guide, each one a blank screenful of nothing. Time on page dropped when I tried it. Readers assume the page is broken.
What worked better was treating slots differently by position. Above the fold and in the first screen after it, reserve exact space, because a shift there is the one that hits the reader mid sentence and the one that scores worst. Further down, let the container collapse to zero and insert the ad with a CSS transform rather than a layout change, or load it only once it is within 400 pixels of the viewport so the shift lands offscreen. Shifts outside the viewport do not count toward CLS, which sounds like gaming the metric and is actually the metric working as intended: nothing moved under anyone's eyes.
One more detail that cost me an afternoon. Multi size ad units are the enemy of reserved space. If a slot can return a 300 by 250 rectangle or a 320 by 100 banner, you cannot reserve correctly for both, and the compromise gap is visible either way. I cut our mobile units to one size each and the remaining ad shifts went away entirely.
The other six sources, ranked by how often they bite
Ads get the blame and account for maybe half of it on a typical guide page. The rest is ordinary markup carelessness, and most of it takes minutes to fix.
| Source | What you see | Fix |
|---|---|---|
| Images without dimensions | Text jumps as each screenshot paints | Set width and height attributes, let CSS scale them |
| Web font swap | First screen retypesets after 300ms | Preload the font, match fallback metrics with size adjust |
| Cookie or consent banner | Whole page drops on first visit | Overlay it with fixed position, never in the document flow |
| Lazy loaded video embeds | Gap collapses then expands | Wrap in a container with aspect ratio 16 / 9 |
| Related guides widget | Footer area pushes content down late | Render it server side, or reserve its exact height |
| Sticky patch notes bar | Content slides when the bar appears | Reserve its height in the body padding from the start |
The font one deserves a note because it is the least obvious and it hit our tier list tables hardest. When a fallback font has different metrics from the real one, every line of text changes height when the swap happens. Chrome now lets you tune the fallback with the size adjust and ascent override descriptors in an at font face block, and getting those within a couple of percent removed a 0.19 shift for us without touching a single ad.
For embeds, the aspect ratio wrapper is the whole fix, and it pairs with the loading approach in our guide to lazy loading YouTube embeds on guide pages. Reserve the box, then fill it whenever you like.
If you only have one evening, do this: add width and height to every image in your post template, move the consent banner to a fixed overlay, and reserve exact space for the single ad slot nearest the top of the page. On our wiki those three changes took the mobile field CLS from 0.31 to 0.08 over the following 28 days, which is how long the report takes to catch up. The other six slots never needed touching.
FAQ
Does CLS only measure the page load?
No, and that assumption is why so many sites think they pass. It runs for the whole life of the visit, so an ad that fills after thirty seconds of reading counts the same as one that fills immediately.
How long until Search Console shows my fix?
The Core Web Vitals report uses a rolling 28 day window of field data, so expect four weeks before the group moves and a few days more for the status to flip. Checking daily for a week will only make you doubt a fix that worked.
Will removing ads from guide pages fix CLS?
It removes the largest source and leaves the rest. Sites that strip their ads and still fail usually have undimensioned screenshots and a font swap, which together are often worth 0.15 on their own.
Is a small layout shift ever acceptable?
Anything the reader caused is not counted at all, so a dropdown that expands on tap is fine. A shift within 500 milliseconds of a real interaction is also excluded. Everything else should be treated as a bug. There are more template level fixes in our web tips section.
More in Web Tips
Web Tips
Last Updated Dates on Guide Pages, Done Honestly
Our wiki had a raid guide that was genuinely current.
Web Tips
How to Fix Discord Link Previews for a Game Site
Someone in our clan Discord posted a link to a raid guide I had spent a weekend writing, and the embed came out as the site...
Web Tips
How to Lazy Load YouTube Embeds on Game Guide Pages
A boss guide with six embedded YouTube clips feels helpful to write.