Lighthouse Core Web Vitals: What It Tests and Misses

Lighthouse Core Web Vitals: What It Tests and Misses

Lighthouse Core Web Vitals pairs two different things. Lighthouse is Google's open-source tool that tests a page in one simulated lab load. Core Web Vitals are three metrics that Google measures on real visitors: Largest Contentful Paint (LCP), Interaction to Next Paint (INP), and Cumulative Layout Shift (CLS). Lighthouse reports lab estimates of LCP and CLS, and it uses Total Blocking Time as a stand-in for INP. It diagnoses problems. It does not decide whether you pass.

That difference explains why a page can score 95 in Lighthouse and still fail the Core Web Vitals assessment in Search Console. This article covers why the two diverge, how to run a Core Web Vitals test in Lighthouse, how to read each metric in the report, what a good score means, and how to confirm your fixes with real-user data.

At Nostra AI we work on ecommerce storefront speed every day, so we see this gap constantly. Teams chase a Lighthouse number while shoppers on mid-range phones experience something else. The sections below show you how to use Lighthouse for what it does well.

Why Lighthouse and Core Web Vitals are not the same thing

Lighthouse is a tool, and Core Web Vitals are a scorecard. Lighthouse runs one simulated page load and reports what it saw. Core Web Vitals are judged on real visits collected from Chrome users over time. Mixing them up is the most common mistake in speed work.

Lab data versus field data

Lab data comes from a controlled test. Lighthouse loads your page on an emulated device with throttled CPU and network, so the result is repeatable and works on staging sites or pages behind a login. The cost is realism. No real person scrolls, taps, or opens your cookie banner during the test.

A laptop cabled to a smartphone on a stand, with a stopwatch and several other phones behind.

Field data comes from actual visitors. Google collects it in the Chrome User Experience Report (CrUX), and the Core Web Vitals assessment uses the 75th percentile of those visits. A page passes only when three out of four real visits are good. Field data shows what happened, but it rarely tells you why.

What each one covers

The table below sets the two side by side.

Lighthouse Core Web Vitals
What it is Testing tool with many audits Three user-experience metrics
Data type Lab, one simulated load Field, real Chrome visits
Loading LCP (lab estimate) LCP
Responsiveness Total Blocking Time as a proxy INP
Visual stability CLS during page load CLS across the visit
Other categories Accessibility, SEO, best practices None
Best use Diagnosing and catching regressions Judging real user experience

Google's ranking systems use the field data. Your Lighthouse score is not a ranking input, but it points at the same problems that hurt your field numbers, such as heavy scripts and slow images.

Lighthouse tells you what to fix. Field data tells you whether it worked.

Why the gap matters

Consider a hypothetical store. Its product page scores 94 in Lighthouse on mobile, so the team moves on. Real shoppers arrive on mid-range phones over busy mobile networks, open a chat widget, and tap filters while scripts still run. Their 75th percentile INP lands at 280 ms, above the 200 ms "good" limit. The page fails the assessment despite a green lab score, and Lighthouse could never have caught it because it does not interact with the page.

The fix is not to ignore Lighthouse. Use it as the debugger and field data as the judge. The rest of this article follows that split.

How to run a Core Web Vitals test in Lighthouse

The fastest way to run a Core Web Vitals test in Lighthouse is PageSpeed Insights. Paste a URL, click Analyze, and read the report in under a minute. Use Chrome DevTools when the page is behind a login or still on staging.

Four ways to run Lighthouse

  • PageSpeed Insights (pagespeed.web.dev): no install, and it shows real-user field data above the lab results when your page has enough traffic.
  • Chrome DevTools: open the Lighthouse tab, click Analyze page load, and wait 30 to 60 seconds. It works on authenticated pages.
  • Chrome extension: the same engine, started from the toolbar.
  • Node command line: built for scripts and continuous integration, and the basis of Lighthouse CI.

The command line suits repeated testing. Install it once, then run it against any URL, as described in Google's Lighthouse documentation.

npm install -g lighthouse
lighthouse https://www.example.com --only-categories=performance --view

Settings that make results trustworthy

A careless run produces a number you cannot compare with anything. Follow these steps for a result you can act on:

A laptop showing a browser developer panel, with a stopwatch and a notepad marked with five tally marks.

  1. Open an incognito window with extensions disabled. Extensions inject scripts and skew the result.
  2. Test in mobile mode first. It is the stricter test and the one most shoppers experience.
  3. Close heavy tabs and apps, then run the test three to five times and keep the median.
  4. Test each template, not only the homepage. For a store that means home, collection, product, and cart.
  5. Record the exact settings, so a later run is comparable.

One Lighthouse run is an anecdote. The median of five is a measurement.

How to read the Core Web Vitals in a Lighthouse report

Lighthouse shows LCP and CLS directly in its Metrics section, and it reports Total Blocking Time (TBT) where INP would be. Read those three numbers first, then open the audits that explain them.

Thresholds for each metric

Google defines "good" for each vital at the 75th percentile of real visits. Learn more about the metrics in Google's Core Web Vitals overview.

Core Web Vital Good Needs improvement Poor In Lighthouse
LCP 2.5 s or less 2.5 to 4 s Over 4 s Reported
INP 200 ms or less 200 to 500 ms Over 500 ms Not measured, TBT shown
CLS 0.1 or less 0.1 to 0.25 Over 0.25 Reported for load

These are field thresholds. Lighthouse colors its lab values in a similar way, and it marks TBT green under 200 ms and red above 600 ms.

Why INP is missing from the lab

INP measures how quickly a page responds when a visitor clicks, taps, or types. Lighthouse loads the page and never interacts with it, so it cannot measure INP at all. TBT adds up the time the main thread was blocked by tasks longer than 50 ms during load. A low TBT usually goes with a decent INP, but it does not guarantee one.

To see real interaction delays, record a trace in the DevTools Performance panel while you click through the page. Field data from a real-user monitoring tool is better still.

What drives the performance score

The score is a weighted blend, and three of its five metrics are Core Web Vitals related. In current Lighthouse versions the weights are:

A bar chart of five Lighthouse metric weights, with LCP, CLS, and TBT together making up 80 percent.

  • Total Blocking Time: 30%
  • Largest Contentful Paint: 25%
  • Cumulative Layout Shift: 25%
  • First Contentful Paint: 10%
  • Speed Index: 10%

LCP, CLS, and TBT make up 80% of the score, so improving them moves the number most.

A Lighthouse score is a weighted blend of lab metrics, not a measure of how your visitors feel.

Audits to open first

Each metric has an audit that names the cause:

  • "Largest Contentful Paint element" shows which element set your LCP time.
  • "Avoid large layout shifts" lists the elements that moved the most.
  • "Avoid long main-thread tasks" lists the scripts behind high TBT.
  • "Reduce the impact of third-party code" shows which vendors block the main thread.

What counts as a good Lighthouse score

A good Lighthouse performance score is 90 to 100, the green range. A score of 50 to 89 needs improvement, and 0 to 49 is poor. Treat the number as a health check, not a goal.

How to interpret the number

Scores move between runs, devices, and networks, so a swing of a few points is noise. Do not chase a perfect 100 by stripping out useful tools, such as a reviews widget or an analytics tag, just to please the audit. Aim for green on LCP, CLS, and TBT individually, which tells you more than the blended score.

The number that matters for search is different. Your page passes the Core Web Vitals assessment when LCP, INP, and CLS are all good at the 75th percentile of real visits. A 98 in the lab with a failing field result is a fail, and a 72 in the lab with a passing field result is a pass.

Are Core Web Vitals still relevant?

Yes. Google documents them as part of its page experience guidance, and Search Console still reports them for your site. They are not a trump card, though. Google says its systems prioritize helpful, relevant content first, so a strong page with average vitals can still outrank a fast, thin one.

The honest reason to fix them is that slow, jumpy pages lose shoppers. Ranking benefit comes second.

Pass the assessment for your visitors first. The ranking benefit follows.

How to fix what Lighthouse finds and confirm it

Fix the largest lab finding first, then verify the change in field data. That verification takes weeks, not minutes, because real-user data is averaged over time.

A practical fix order

Work down this list on each template:

  1. Server response time. A slow first byte delays every other metric, including LCP.
  2. LCP image. Preload the hero image if the browser discovers it late, and never lazy-load it.
  3. Image dimensions. Set width and height on images and video so the browser reserves space and CLS stays low.
  4. Third-party scripts. Defer what you can, and replace heavy embeds with a lightweight facade that loads on click.
  5. Long tasks. Split or delay scripts that block the main thread for more than 50 ms.

Confirming the result in field data

CrUX uses a 28-day rolling window and judges the 75th percentile. After a fix, your faster sessions are a small slice of that window, so the score barely moves in the first week. Expect two to three weeks before the improvement shows. Check the field section in PageSpeed Insights and the Core Web Vitals report in Search Console, and do not judge a fix from a Lighthouse rerun alone.

Where an edge layer helps

For ecommerce stores, the first item on the fix list is often the hardest to change without engineering time. Dash, Nostra AI's Edge Delivery speed agent, caches a store's HTML at the edge and goes live with a DNS-level change, with no code changes. It improves Time to First Byte by 5 to 10x, and customers' Core Web Vitals typically improve within 14 to 28 days. Time to First Byte is not a Core Web Vital, but a faster first byte gives LCP a head start.

Fix what the lab shows, then wait for the 28-day window to prove it.

Use Lighthouse to diagnose, field data to judge

Lighthouse and Core Web Vitals work best as a pair. Lighthouse gives you a repeatable lab test that points at the cause of slow loading, layout shifts, and blocked scripts. Core Web Vitals field data tells you whether real visitors are better off. Run Lighthouse several times per template, fix what it flags, and let the 28-day window confirm the result.

If bot traffic is also inflating your sessions and muddying the analytics you use to judge speed work, Edge Protect filters it before it reaches your store.