PageSpeed-Insights Reports High CLS when Nothing is Shifting?

Question

I've had low CLS on Google PageSpeed-Insights (mobile screen size) for a long time but recently its jumped quite high.

I cant find any shifts in the page loading pictures.

Chrome Lighthouse test shows low or zero CLS and Google Search Console shows no issues.

I'm expecting to be able to find the layout shift reported in PageSpeed-Insights but I cant find any layout shift.

Can Google PageSpeed-Insights be unreliable ??

Answer 1

Actually it looks like I can find the shift under the small section :

Show audits relevant to: All FCP LCP TBT CLS

and picked CLS

The shift is below the screen size used in chrome lighthouse but shows up in PageSpeed Insights and looks to be some text reformatting.

Thanks for your assistance.

Stephen

Answer 2

Can Google PageSpeed-Insights be unreliable ??

Of course any software can have bugs, but every time I've seen this brought up, the shifts can be found on the page.

It's important to realise a few things:

First up Page Speed Insights shows real user experience from actual people using Chrome on your site. This is shown in the top section.

It's important to look to see if you are seeing URL data, or it's falling back to origin-wide data, as URL-specific data is not available.

This will help you understand if the issue is with your page, or your site in general.

The CLS in this section is for the lifetime of the page and not just page load. If you have CLS on scroll, then this will show here, but not in the lab-based test beneath which just does a load of the page without any interaction.

Similarly Single Page Applications (SPAs) can often have CLS as the user navigates the app, that will not be seen on the simple page-load Lighthouse that PSI uses.

Beneath this "Discover what your real users are experiencing" section is the lab-based Lighthouse test labelled "Diagnose performance issues". This runs a page load Lighthouse test under simulated conditions that may or may not be representative of actual visitors conditions when they visit your site.

It may be most of your visitors are on faster mobiles or network conditions than PSI uses here. Or most of your visitors are repeat visitors that already have some of the page cached.

The point of this section, is that Lighthouse can run some extra tests, that take a bit of time (you'll notice this section takes longer to load than the real user data which is a simple look up and so loads quickly). This can give some helpful extra information and suggestions for things that might improve the performance of your page. However, you should always check the Lighthouse lab data back to the Real User field data and act appropriately. If the field data says you have CLS but the lab-based data does not, then you almost certainly do but need to identify some post-load CLS issues.

So the big questions you need to answer are:

1. Is it the field data or lab-based data that is showing the CLS in PSI? Or both?
2. Is it url or origin-level data?

Chrome Lighthouse test shows low or zero CLS and Google Search Console shows no issues.

That is interesting as usually you'd expect one of those to match, but it really depends what data you want to compare this too from PSI, which is unclear from your question.

Chrome Lighthouse test shows low or zero CLS

If Chrome Lighthouse test shows low or zero CLS but PSI Lighthouse shows CLS then this suggests you are not loading the page in the same way. Common issues here include:

• you having dismissed a cookie banner as you use the site all the time, but PSI sees the cookie banner
• you have some assets cached or downloaded (like Fonts) that PSI does not. Try ticking the "Disable cache" option in Network panel.
• you are serving different results to different countries and PSI is seeing different content
• you are using a faster connection or different screen size than PSI so loading happens differently

Google Search Console shows no issues

If GSC shows no issued but the real user experience section in PSI is showing a difference, then this is possible for a few reasons:

• GSC groups pages into page groups (e.g. product pages, blog pages...etc.). So it's possible some of your pages have CLS issues, but others in that group are offsetting it enough so the group passes and shows no issues.
• It's possible that no-index pages are being included in PSI's origin-level data but not GSC's data

Hopefully that gives you some more information to go on, but if still confused, then you need to edit your question to explain exactly where in PSI you're seeing the CLS issue and, ideally, provide an example URL. But I would be very surprised if this is PSI being unreliable and it's more likely that you have a real CLS issue under certain circumstances.