Lighthouse Score Good but Search Console Needs Improvement: Here’s Why

A good Lighthouse score and a Search Console warning can both be true at the same time because they measure two different things. Lighthouse runs a simulated test on a single visit. Search Console reports real visitors over the past 28 days. A page can pass the simulation and still fail for actual people on actual phones.

You ran Lighthouse. Green across the board. Maybe even a perfect 100. You felt good about it, closed the tab, and went on with your day.

Then you opened Search Console and saw it. Needs improvement. Sometimes poor. The same site. The same pages. Two completely different verdicts.

If you’re like most business owners I talk to, your first thought is that something is broken. Your second thought is that maybe Google just hates your website. 

Neither is true. What’s actually happening is simpler and weirdly more useful once you understand it, because it tells you exactly where your real users are struggling, not where a robot in a lab thinks they might.

This post covers what Lighthouse and Search Console each actually measure, why INP is probably your real problem, what that confusing 28 day window means for when you’ll see results, and why one slow page can drag down a whole group of fast ones. 

By the end you’ll know exactly which number to trust and why.

Lighthouse tests a robot. Search Console tracks your actual visitors.

Here’s the part nobody explains clearly enough. Lighthouse doesn’t watch real people use your site.

It opens your page once, on a simulated device, on a simulated network connection, and times how everything behaves under those fixed conditions. 

It’s a controlled experiment. Same inputs every time, which is exactly why developers love it for catching obvious problems before launch.

Search Console doesn’t run a test at all. It pulls from something called CrUX, the Chrome User Experience Report, which quietly collects real performance data from actual Chrome users who land on your pages. Different phones. 

Different connections. Someone on fiber in Manchester and someone on a patchy 4G signal in a different city entirely.

That’s the data Google actually uses to judge your page experience for ranking purposes.

Lighthouse tests a robot. Search Console tracks your actual visitors. One is a rehearsal. The other is opening night, every night, for 28 days straight.

This is why Google has said plainly that Lighthouse scores aren’t used for search ranking at all.

They’re diagnostic. Field data from CrUX is what counts.

If you’ve been polishing your Lighthouse number and ignoring Search Console, you’ve been optimising for the wrong audience the whole time.

Why INP is quietly wrecking sites that look fine

If you’re failing Search Console despite a good Lighthouse score, INP is usually the culprit, and it’s the metric most business owners have never heard of.

Interaction to Next Paint measures how fast your page responds when someone actually clicks or taps something. Not how fast it loads. 

How fast it reacts once a real person tries to use it. A good score needs at least 75% of a page’s interactions to respond within 200 milliseconds. Anything slower starts dragging your rating down.

Here’s why Lighthouse misses this so often. According to the 2025 Web Almanac, 77% of mobile pages now achieve a good INP score globally, which sounds reassuring until you check the same data for high traffic sites. 

Among the top 1,000 most visited websites, only 53% pass. The busier and more feature rich your site, the more JavaScript you’re running, and the more likely a real visitor hits a slow, sticky interaction that Lighthouse’s single simulated test never bothers to recreate.

Think about what actually happens on a real visit. Someone opens your contact form. They click submit. There’s a popup chat widget loading in the background, an analytics script firing, maybe a third party booking calendar competing for the same processing thread. 

The click registers a half second late. Lighthouse never clicked anything, so it never saw that. Your real visitor felt it.

Your site can look instant in a lab and feel sticky in someone’s hand. That gap is INP.

The 28 day window that makes your fixes feel invisible

This is the part that frustrates people most, so let’s be straight about it. Ship a fix today, and Search Console will not reflect it today. Or tomorrow. 

The Core Web Vitals report works on a rolling 28 day window of real user data, which means your current rating is really an average of the last month of visits, not a snapshot of right now.

Fix a slow loading image today, and for the next several weeks your report is still partly haunted by the visitors who experienced the old, slower version. The new, faster experience needs to accumulate enough fresh data to outweigh the old data before the needle moves. 

Most site owners give up right around week two, convinced their fix didn’t work, when really it just hasn’t had time to show up yet.

I tell every client the same thing when we ship a performance fix. Mark the date. Set a reminder for four weeks out. Don’t even open the Core Web Vitals report before then, because you’re not going to see the truth yet, and a false flat line will make you doubt work that’s actually succeeding quietly in the background.

This is also exactly why Search Console and a live Lighthouse test can disagree the week after a fix ships. Lighthouse already sees the improved code. Search Console is still digesting a month of history. They’re not contradicting each other. They’re just reporting on different timeframes.

One bad page can sink a whole group of good ones

Here’s a detail that catches almost everyone off guard, including some developers. Search Console doesn’t always report on individual URLs. 

It groups similar pages together, your blog posts, your product pages, your category pages, and assigns the whole group a single status based on its worst performing member.

So if you have forty blog posts and thirty nine of them load beautifully, but one has an oversized hero image or an embedded video that chokes on mobile, that one page can drag the entire group’s status down to needs improvement. 

You check ten of your best posts manually, they all look fine individually, and you’re left baffled about why the report still says there’s a problem.

The fix isn’t to panic about your whole blog. It’s to find the page actually dragging the average down. Search Console’s report shows example URLs for each flagged group specifically so you can identify the outlier rather than guessing. 

Hunt that page down, fix what’s slow on it, and the group’s status improves once enough fresh data rolls in.

This single mechanic explains a huge share of the confused messages I get. The site isn’t broadly broken. One page is quietly sabotaging thirty others that were never the problem.

What to actually check, in the right order

Stop refreshing Lighthouse. Here’s the order that actually gets you somewhere.

Open Search Console’s Core Web Vitals report first. It’ll tell you which URL groups are flagged and which specific metric is failing, LCP, INP, or CLS. That’s your real starting point, not a guess.

If INP is your problem, look at what’s running on your page beyond your own content. Chat widgets, analytics tags, booking calendars, ad scripts. Heavy JavaScript execution and third party scripts are the most common causes of poor responsiveness, and they’re often tools you added for good reasons that are now quietly taxing every click.

If it’s LCP, the loading metric, check your largest image or block of text on the page. It’s almost always an oversized hero image or a render blocking resource sitting in front of it.

Then wait. Genuinely wait the 28 days before judging whether your fix worked. And once you’ve fixed the obvious culprit, check whether it’s dragging down a whole URL group rather than assuming the fix only helped one page.

Final thought

Your Lighthouse score was never lying to you. It was just answering a different question than the one Google’s actually asking. Lighthouse tells you how a robot experienced your page once, under perfect lab conditions. Search Console tells you how real people, on real phones, with real distractions and real patience limits, experienced it over the last month.

That gap between the two isn’t a glitch in your setup. It’s the entire point of having two different tools. One catches obvious problems early. The other tells you the truth that actually decides your rankings.

So the next time they disagree, you’ll know exactly why, and exactly which one to trust.

Not sure which number is actually hurting your rankings?

If you’d like someone to look at your actual Search Console data and tell you honestly what’s dragging your scores down, I’d love to help. Send me a message and I’ll walk through it with you.

Frequently Asked Questions