PageSpeed Insights vs Search Console: Why They Disagree
Your URL passes in PageSpeed Insights while Search Console flags the group, or PageSpeed Insights fails while Search Console is green: why they differ and how to find the slow pages.
PageSpeed Insights vs Search Console in short
PageSpeed Insights and Search Console read the same Chrome UX Report (CrUX) data: the 75th percentile over 28 days. PageSpeed Insights shows one URL, or your whole site when that URL has too few visits. Search Console rates groups of similar indexed URLs by their worst metric and counts URLs. One page can pass in one tool and fail in the other.
Do PageSpeed Insights and Search Console use the same data?
Yes. The field section at the top of PageSpeed Insights and the Core Web Vitals report in Search Console both read CrUX at the 75th percentile (p75) of the last 28 days, split into Mobile and Desktop, with the same limits for good: Largest Contentful Paint (LCP) up to 2.5 s, Interaction to Next Paint (INP) up to 200 ms and Cumulative Layout Shift (CLS) up to 0.1. They differ in which pages they put together and in what they count.
The 0 to 100 Performance score lower down in PageSpeed Insights, and the Lighthouse metrics under it, come from one simulated load. Search Console never reads them (why Lighthouse and field data differ). The dates differ a little too: in PageSpeed Insights the 28 days end about two days ago, in Search Console a few days earlier (how the 28-day CrUX window moves).
Does PageSpeed Insights show This URL or Origin?
PageSpeed Insights shows the page you entered under "This URL" only when that URL is publicly discoverable and has enough Chrome visits of its own. When the page is new, gets too few visits or is not discoverable, the field section switches to "Origin": the 75th percentile of all page views on the whole site (protocol, host and port). When the two tools disagree, that switch is the first thing I check.
The numbers on this page come from the CrUX API, the same data both tools read. I queried it on 22 November 2026 (collection period 24 October to 20 November 2026), on phones unless I say desktop. The origin www.theguardian.com has an LCP of 937 ms, INP 111 ms and CLS 0.04. Every page I measured that has its own record differs from it. Five section fronts, such as /football and /politics, load faster: 682 to 827 ms LCP and 0.00 CLS. Five articles from October run from 893 to 1,027 ms LCP, four of them slower than the origin, with CLS from 0.03 to 0.07.
Even on a news site that size, some pages have no record of their own. Two articles published on 21 November had none, because the collection period ended on 20 November: every new page starts without URL data for its first days. An information page from January 2016 had too few visits. With too little data, PageSpeed Insights falls back to the origin and shows the whole site's numbers, as in the second screenshot below.

This URL: an October article with its own data

Origin: a new article gets the whole site's numbers
PageSpeed Insights, Mobile, two articles on www.theguardian.com, captured on 22 November 2026.
Origin also holds pages Search Console never shows. Per the CrUX methodology, the origin's data adds up the visits to all pages on that origin, including pages that are not publicly discoverable themselves. Search Console's report holds only indexed URLs. So cart, checkout, account and internal search pages can count toward Origin in PageSpeed Insights and never appear in Search Console.
The home page is where this catches people out. What I see all the time: someone pastes the home page into PageSpeed Insights, sees green numbers under "Origin" and assumes the whole site passes, while Search Console flags a template. On a content site or a store, search visitors land on articles or product pages, so the home page itself can have too few visits and switch to Origin. In a thread in the CrUX forum, a site owner tested their home page and got a passing Origin LCP of 2.1 s (79% good), while Search Console flagged their blog group at 4.1 s. Barry Pollard of Google explained it: Origin weights every page by its page views, so a fast, popular group of pages pulls the origin into the green while the blog fails.
When the home page does have its own data, it is still one page. The Guardian's /uk home page has an LCP of 776 ms and CLS 0.00, both better than the origin, but an INP of 135 ms against the origin's 111 ms: the only page I measured with a slower INP than the origin.
Why does Search Console flag a URL group when the page I test passes?
Search Console puts indexed URLs with a similar user experience in one group, takes one 75th-percentile value per metric for that group, and gives every URL in it the status of the group's worst metric on that device. A page that is fast on "This URL" in PageSpeed Insights can sit in a group whose slower pages push the group's p75 over the limit. A group with good LCP (up to 2.5 s) and good INP (up to 200 ms) is still Poor for every URL in it when the group's CLS is over 0.25, and Needs improvement when it is over 0.1. That makes the report stricter than Core Web Vitals as a Google ranking signal.
Click an issue in the Core Web Vitals report and then its URL group, and Search Console shows one example URL, the number of similar URLs and the group's p75: one page standing for a group, usually one template. As Google's help page says, the numbers for a specific URL in PageSpeed Insights might not match its group. Even when that example URL has its own data, you are looking at one page of many. Five Guardian articles from October run from 893 to 1,027 ms LCP on phones, against 937 ms for the origin.
The counts differ too. Search Console counts URLs, PageSpeed Insights shows shares of page loads. As Rick Viscomi of Google put it in the Chrome UX Report forum, Search Console reports the number of pages rather than a percent of experiences, so the proportions don't line up. A quiet template with many indexed URLs can add a tall column of poor URLs in Search Console and hardly move the Origin percentages in PageSpeed Insights, which are weighted by page views.
URL parameters split the two tools as well. Search Console assigns data to the actual URL instead of the canonical URL, and keeps URL parameters when it tells pages apart. PageSpeed Insights strips all parameters, and CrUX strips query strings and fragments. When I asked the CrUX API for a Guardian football article of 28 October with ?utm_source=email added, it returned the record of the clean URL: 1,027 ms LCP, 101 ms INP and 0.06 CLS.
Why is PageSpeed Insights red when Search Console is green?
PageSpeed Insights can show "Core Web Vitals Assessment: Failed" or a red score while Search Console has no failing URLs. The reason is in one of four places on screen:
- The lab score. Look at the field section at the top first. When the assessment passes there and the Lighthouse score below it is low, your visitors pass and only the simulated load struggled. Reading the lab score, often on the Desktop tab, as the Core Web Vitals result is a mix-up I see all the time.
- Mobile against Desktop. Match the tab in PageSpeed Insights to the report in Search Console. On www.theguardian.com the origin's CLS is 0.04 on phones and 0.12 on desktop, over 0.1. Two October articles go from 0.06 and 0.04 on phones to 0.15 on desktop.
- Unindexed pages inside Origin. When the field section says "Origin", slow cart, checkout, account and search pages can count toward it, and Search Console shows only indexed URLs.
- A change of the last few days. Compare Search Console's "last updated" date with today: a change shows in PageSpeed Insights a few days before Search Console catches up.
Why does Search Console say there is not enough data?
Search Console shows "No data available" in the first few days after you create a property, or when CrUX has too little data on that device for your indexed URLs. A URL group without enough data for both LCP and CLS is left out of the report. Both tools read CrUX, and each has its own minimums and fallbacks:
| Rule | PageSpeed Insights (field section) | Search Console (Core Web Vitals) |
|---|---|---|
| Fallback order | "This URL", then "Origin" (all pages on that origin), then no field data | URL group of similar pages, then a higher-level origin group, then left out |
| Which URLs count | Publicly discoverable URLs for "This URL"; all pages on the origin in "Origin" (indexed or not) | Indexed URLs only (not a complete list of all indexed URLs) |
| Metrics required | Both LCP and CLS for the assessment (assessed without INP when INP lacks enough samples) | Both LCP and CLS for the group, or the URL is left out of the report |
| Device split | Assessed per device on the Mobile and Desktop tabs | Separate Mobile and Desktop reports; one can show groups while the other says "No data available" |
| New page or property | A newly published URL falls back to "Origin" until it collects enough visits | A new property can take a few days before CrUX data appears |
So a quiet site can have numbers in one tool and none in the other.
How do you find the slow pages in a Search Console URL group?
Open the flagged URL group in Search Console, test its example URL and three or four popular pages of the same template in PageSpeed Insights on "This URL", follow the pages with their own data in the CrUX History report, and use Real User Monitoring (RUM) for the pages that switch to "Origin".
- Open the issue and the URL group in Search Console. Pick Mobile or Desktop to match the failing report, click the Poor or Needs improvement issue, then the URL group. You get the failing metric, the group's p75, the number of similar URLs and the example URL, so you know which template and which metric to test.
- Test the example URL and three or four popular URLs of that template in PageSpeed Insights. Use the same device tab and check that the field section says "This URL". Popular pages are the most likely to have their own CrUX record. Compare each page's p75 for the failing metric with the group's p75 to see which URLs pull the group over the limit.
- Check the trend in the CrUX History report. Put each URL that had its own data into the CrUX History report and read its weekly bars.
- Use RUM for the URLs that fall back to Origin. When most pages in the group switch to "Origin", CrUX gives you only the site total, while RUM measures each page on its own. Even on www.theguardian.com, articles published on 21 November and a quiet 2016 page had no URL record on 22 November. Only measuring in the browser on your own pages gives every URL in the group its own number.
How I check this in real-visitor data
When a Search Console group fails and PageSpeed Insights keeps switching to "Origin", I open CoreDash, filter on Mobile and the template's path (such as /blog/ or /products/), and sort the URLs by p75. CoreDash, the Real User Monitoring tool I built, records every page view, so every page in the group gets its own p75 per device and path, including the URLs without a CrUX record.
Real-visitor monitoring shows different numbers from CrUX again, for other reasons (why CrUX and RUM show different numbers).
Search Console still flags a URL group while every page you test in PageSpeed Insights passes or switches to Origin? Bring your Search Console report to an hour with me on your field data. We go through your groups, find the template and the element that pull your p75 down, and decide what your developer fixes first.

Need a faster site?
I'm Arjen, and I've been fixing slow sites for 17 years. If you're running into issues you can't fix yourself, send me your URL and let's have a chat.
Send me your URL
Why isn't our fix working?
Sometimes the CDN still serves the old HTML, or the fix went into the desktop template and mobile uses another one. Send me your URL and the problem, and I reply with a few times. I check your field data first, then we spend one hour on your site, and within a day you have my notes in writing. €350 excl. VAT.
Send me your URLNo developer to build the fix? I do the work myself, on a copy of your site.