PageSpeed Insights gave your homepage 95 out of 100. Then you opened Google Search Console and found the same pages listed as slow. Two reports, both from Google, opposite verdicts. It feels like one of them must be broken. Neither is.
The two tools answer different questions with different evidence. Know which question each one asks, and the disagreement stops being confusing. Working out which number to trust is the first step towards a site that feels fast.
What the lab score really measures
PageSpeed Insights runs your page in a controlled test: one simulated device, one connection speed, one location, an empty cache, and nobody clicking anything. Google calls this lab data, meaning results gathered in a fixed environment with set conditions, so the test behaves the same way every time.
Because hardly anything changes between runs, you can test before and after a change and see what your work achieved. What you cannot do is treat the 0-100 number as a verdict on your visitors' experience.
Run the same page twice in an afternoon and the score can move by several points: a script that answers more slowly the second time, a moment of congestion on the test connection. The number summarises one run of one test. It is not a measurement of your customers.
What Search Console measures instead
Search Console's speed report is built on field data: what really happened for actual people who visited your site over the previous 28 days, drawn from Chrome's browsing statistics across millions of visits. No simulation, no single test machine. It includes old phones, patchy mobile connections, visitors halfway through a video call, and people returning to the same page for the fourth time.
Three measurements decide the verdict. Largest Contentful Paint, how long the main content takes to appear, has to come in within 2.5 seconds. Interaction to Next Paint, how quickly the page reacts when someone taps or clicks, has to be 200 milliseconds or less. Cumulative Layout Shift, how much the page jumps around while loading, has to stay under 0.1. Google judges each at the 75th percentile, so three visits out of four must meet the target, and mobile and desktop are graded separately.
Worth knowing: the loading target is still 2.5 seconds, not the 2.0 a number of 2026 blog posts claim. Google's own guidance lists 2.5 seconds for loading, 200 milliseconds for responsiveness and 0.1 for stability.
Why the two reports disagree
The gap between them is not noise. Each difference points at something you can act on.
Devices and connections. A lab test uses one device profile at one speed. Your audience is a spread of them, from current phones on fibre to five-year-old handsets on hotel wi-fi.
Location and cache. Search Console combines data from every country your visitors arrive from; a single test run never sees that. Lab tests also load the page cold. Repeat customers get a warm, partly cached page that loads faster than any fresh test.
Behaviour. A lab test stops once the page has loaded. Real people scroll, tap and type, and that is exactly when layout shifts and slowdowns tend to appear.
What the page shows. Personalised content or an A/B test variant blends into the field figures, while your lab run loads one version only. If the slow variant is the one the test never sees, the two never reconcile.
Which one to believe
Believe the field data about the problem, and use the lab test to find the cause. Search Console is telling you what your visitors experienced; PageSpeed is telling you how a controlled version of the page behaves, so you can debug it.
If Search Console flags a page as poor while a lab run scores 95, your customers are having a worse time than the test suggests, and their experience is the one that matters. Trust the field. A strong lab score still says the page performs well under good conditions, which points the finger at a real-world condition rather than the design: a heavy script firing when someone taps a button, an image that crawls on a weak connection, hardware that struggles.
Why "no data available" is not a failure
If your report says no data is available, the likeliest explanation is a brand-new Search Console property or not enough visits yet for that device type. Google needs a threshold amount of real-world information before it publishes figures, and it grades groups of similar pages together rather than every address one by one.
No score is not a failing score. A new site with little traffic should use lab tests as a stand-in and revisit the field report once visitor numbers build up.
What to fix first
A simple order of operations saves weeks of guessing.
1. Let Search Console pick the target. Whichever of the three measures is failing is your priority, not whichever one a testing tool highlighted. If the report is happy, resist the urge to polish a number nobody is complaining about.
2. Reproduce it in a lab run. Open PageSpeed Insights on the same page, switch to mobile, and read the list of specific problems rather than the headline score.
3. Fix the biggest single thing first. Oversized images, scripts that block the page from rendering and third-party widgets that load far more than they need are behind most small-business slowdowns. One meaningful change beats five small ones.
4. Re-test, then wait. The lab number can improve within minutes. The field report fills up gradually, because it covers a rolling window of the previous 28 days, so a fix needs roughly a month to show fully in Search Console. Our Core Web Vitals fix guide walks through the usual culprits in order.
Don't chase the lab number every day. Small swings between runs are normal. Make one change, note the date, and judge it on whether the real-world figures move once the 28-day window has turned over.
One new report you can ignore for now
On 13 October 2026 Chrome started publishing experimental versions of these same three measures that also count in-page navigation on single-page apps, the kind of site that swaps content without reloading. Google says the stable scores stay unchanged while these experiments run, with graduation aimed for May 2027, and only single-page apps with heavy in-page clicking will see a difference. A normal multi-page small business site should expect nothing to change.
Speed shapes whether a visitor stays long enough to enquire, and how your pages perform in search. The reasons speed matter are a good place to start; then read the two reports side by side, one for what to fix and the other for how.
Want a proper read on your site's speed?
Our AI Website Audit checks your pages against the three measures Google actually uses, on real mobile conditions, and gives you a plain-English list of what is slowing the site down and the order to fix it in. Prefer someone else to keep it fast? Our AI-powered hosting includes ongoing performance care as standard.
Get Your AI Website AuditOr talk to the team about a faster website.