CrUX vs RUM: why the numbers don't match
PageSpeed Insights shows one LCP for your page, your real user monitoring another. They count different visitors over different days, and once you line them up they come close.
Why doesn't my RUM match PageSpeed Insights?
In short
- PageSpeed Insights shows Chrome users on desktop and Android who share usage statistics, over a 28-day window, and for a page with too few of those visits the whole site. Your RUM tool counts every browser your script runs in, including the ones CrUX never sees, up to today.
- Compare Chrome with Chrome, the same device and the same 28 days, both at the 75th percentile, and the two numbers come close.
- CrUX is how Google judges your site, so check it first. Your RUM shows which page to fix and whether the fix worked.
Your RUM and PageSpeed Insights count different visits over different days, and each is right about the visits it counts. PageSpeed Insights shows CrUX, the Chrome User Experience Report that Google collects from Chrome users. The two differ in who gets counted, which days, which URL and which kinds of page view, and each of those has a filter. With the filters set, the two numbers line up.
The performance score in PageSpeed Insights is a Lighthouse lab test of a single page load. The part to compare with your RUM is the field data: LCP, INP and CLS at the 75th percentile over 28 days, which is CrUX.
Which visitors does CrUX count?
CrUX counts Chrome users on desktop and Android who share usage statistics with Google, on public pages that get enough of those visits. Your RUM counts every visitor whose browser runs your script, iPhones included.
Sharing usage statistics means usage statistic reporting is on, the browsing history is synced and there is no sync passphrase. That is Chrome on Windows, macOS, Linux, ChromeOS and Android. Safari, Firefox, Edge and the other browsers built on Chromium stay out, and so does Chrome on an iPhone, which runs on WebKit, the engine under Safari. The rules for which pages get in are in the not enough data guide.
Since Safari 26.2 (December 2025) Safari reports LCP and INP, so your iPhone visitors are in your RUM numbers for those two. CLS is still measured only in Chromium browsers, so a RUM CLS comes from Chrome, Edge and their relatives.
In CoreDash you see the split in Explore: pick LCP and set "Split by" to Browser. Every row other than Chrome is traffic CrUX doesn't have. Chrome on an iPhone sits in the Safari row, which is where it belongs for this comparison, since CrUX doesn't count it either. On a site with a lot of iPhone traffic, mobile is where CrUX and your RUM differ most.
| Browser | Page views | p75 LCP | |
|---|---|---|---|
| Chrome | 18,420 | 2.14 s | in CrUX if opted in |
| Safari | 9,860 | 1.88 s | never in CrUX |
| Edge | 1,240 | 2.31 s | never in CrUX |
| Firefox | 610 | 1.96 s | never in CrUX |
| Opera | 180 | 2.62 s | never in CrUX |
LCP split by browser: only the Chrome row can reach CrUX.
Why does my RUM show today while CrUX barely moves?
CrUX data is about two days old, but it is one number over a rolling 28-day window that Google moves one day at a time. Your RUM has today's page views within minutes.
A fix that went live today is one day out of 28 once CrUX reaches that day. Two weeks later the number is still half the old site, and only after 28 days is the whole window new. The 28-day delay myth walks through it day by day.
In your RUM, set the period to today and look at the page views since the release on their own. The CrUX page in CoreDash adds a column for this, "When CrUX turns good": for a vital that isn't good yet, it gives the day CrUX should get there if your last 7 days of Chrome visitors hold.
Why does PageSpeed Insights show the whole site instead of my page?
PageSpeed Insights shows the whole site when your page doesn't have enough CrUX data of its own. It then switches to Origin: every page on the same protocol and host as one number, so all of https://www.example.com together.
That origin number is weighted by traffic. Your busiest pages count the most, so a slow checkout can sit behind a fast home page and a fast blog.
CrUX also merges URLs before it counts them. It drops the query string and the fragment, so a visit to /shoes?utm_source=newsletter counts as /shoes. A page with noindex never gets a CrUX number of its own, but its visits still count toward the origin.
Your RUM has a p75 for every URL that got visits, including the ones CrUX never publishes. On the CrUX page in CoreDash you add up to 20 of your own URLs to the card "Your pages in CrUX", and it shows Google's p75 for each one that has a CrUX record. Then you compare a page with the same page instead of with the whole site.
There is insufficient real-user data for this URL. Falling back to aggregate data for all user experiences on this origin (https://acme-crux.com) instead.
Latest 28-day period · Various mobile devices · Full visit durations · Various network connections · Many samples (Chrome UX Report) · All Chrome versions
PageSpeed Insights on mobile: the field data is CrUX, and without enough data for this page it shows the whole origin.
Does CrUX count back button visits and prerendered pages?
Yes, CrUX counts a page restored from the back/forward cache (bfcache) and a prerendered page as page views, and both usually beat a normal page load by a wide margin.
A restore comes back from memory, and a prerendered page was loaded before the visitor clicked. On corewebvitals.io, from 23 October to 19 November 2026, prerendered pages had an LCP p75 of 0.22 seconds against 0.96 seconds for a normal page load. CrUX mixes those fast views into its one number. Route changes in a single-page app are not page views in CrUX: it adds them to the page view the visit started on.
Your RUM shows each kind as its own row. In CoreDash, open the LCP page and pick "Navigation type" under More: "Page load", Reload, Back/forward, "Back/forward from bfcache" and Prerendered each get their own p75.
A RUM report that leaves restores and prerenders out comes out slower than CrUX, for a reason that has nothing to do with your site. Keep them in when you compare your RUM with CrUX, and split them out when you want the normal page loads on their own.
| Navigation type | Page views | p75 | Over 2.50 s |
|---|---|---|---|
| Page load | 21,300 | 2.24 s | 4,470 |
| Reload | 1,620 | 1.42 s | 160 |
| Back/forward | 640 | 1.65 s | 80 |
| Back/forward from bfcache | 1,180 | 0.52 s | 50 |
| Prerendered | 2,140 | 0.31 s | 40 |
LCP by navigation type: back button restores and prerendered pages are their own rows in CoreDash. CrUX counts them inside its one number.
How do I compare CrUX and my RUM like for like?
Filter your RUM down to what CrUX counts: Chrome only, the same device, the same 28 days as the CrUX record, at the 75th percentile. Then the two numbers come close.
Set the browser to Chrome, which takes out Safari, Firefox, Edge and Chrome on the iPhone. Pick the same device in your RUM as the tab you read in PageSpeed Insights, Mobile or Desktop. For the days, a CrUX record runs from a first to a last day (the collection period in PageSpeed Insights), and the last day is about two days ago. Use exactly those days: the last 28 days counted from today are two days off. Compare a URL with the same URL, or the origin with all your pages. Leave restores and prerenders in, since CrUX counts them too, and read the 75th percentile on both sides.
CoreDash's CrUX page does all of that for you. It reads the CrUX record for your site and puts it next to "Your RUM p75, Chrome": your own Chrome visitors, on the device you picked, over the record's own first to last day.
| Metric | CrUX p75 | Your RUM p75, Chrome | When CrUX turns good |
|---|---|---|---|
| LCP | 0.74 s | 0.89 s | Good in CrUX |
| INP | 45 ms | 44 ms | Good in CrUX |
| CLS | 0.00 | 0.00 | Good in CrUX |
www.corewebvitals.io passes Core Web Vitals in CrUX. Search Console's Core Web Vitals report reads these numbers.
CoreDash's CrUX page for www.corewebvitals.io: Google's p75 next to the p75 of the site's own Chrome visitors over the same 28 days.
On my own site, Chrome against Chrome over the same 28 days, INP came out at 45 ms in CrUX and 44 ms in CoreDash, CLS at 0.00 in both, and LCP at 0.74 and 0.89 seconds, both well inside good.
When your own two numbers still differ after this, these are the usual reasons on any site:
- CrUX has only the Chrome users who share usage statistics, a part of your Chrome visitors. Your RUM has all of them.
- Some browsers built on Chromium, such as Brave, and the browser inside an Android app call themselves Chrome in their user agent. A RUM tool that reads the user agent can count them as Chrome, while CrUX leaves them out.
- On a page or a device with few visits, the p75 jumps with every handful of slow ones, on both sides.
- An LCP image on another domain: without a
Timing-Allow-Originheader a script on the page gets a slightly rounded paint time for it, and CrUX the exact one. - Iframes: CrUX sees what happens inside an embedded frame, a video player for example, and a script on your page doesn't.
The workaround people try first is making the numbers match: switch the RUM to the median, or to 7 days, until it lands on the CrUX value, and call the tool with the higher number broken. All that does is hide the visitors and the days that differ. Compare at p75 over the CrUX record's own days and look at what is left.
Which number is right, and which one should I act on?
CrUX and your RUM are both right, for different questions. CrUX is how Google sees your site. Your RUM tells you what to fix and whether the fix worked.
CrUX is the number behind the Core Web Vitals assessment in PageSpeed Insights and behind the Core Web Vitals report in Search Console, so it is what Google and your SEO team read. Check it first. I do that on every site, also on sites that already run CoreDash.
Your RUM is what you fix with. It names the page, the LCP element or the release that made a number worse, and it shows today whether your fix worked, weeks before CrUX has caught up.
If you own the site, report CrUX to clients and to SEO, and give your developer the RUM to work from. If you are the developer, work in the RUM, and when someone asks why the numbers differ, set your RUM to Chrome, their device and the days of the CrUX record, and show both side by side. For scripts and AI assistants, the get_crux tool in CoreDash's CrUX API returns the CrUX record next to your own Chrome p75 over the same days.
Whether the free CrUX data is enough on its own is on CrUX and Search Console against CoreDash. To look up any site's CrUX history, use the CrUX History tool on corewebvitals.io.
Sources: CrUX methodology, CrUX metrics, Why is CrUX data different from my RUM data?, About PageSpeed Insights, LCP and INP are now Baseline Newly available.