Core Web Vitals and technical SEO: what to check, in what order

Core Web Vitals are three metrics Google uses to measure whether your site feels fast and stable: LCP, INP and CLS. They are not an enormous ranking factor, but they decide how many people stay. And when your content and a competitor’s are level, they are the tiebreaker.

Careful with the guides you find: FID no longer exists. Google replaced it with INP in March 2024. If an article or an agency is talking to you about FID in 2026, they are working from old information.

What are the three metrics and what values should you hit?

Metric	What it measures	Good	Needs work	Poor
LCP (Largest Contentful Paint)	How long the largest element takes to paint	≤ 2.5 s	2.5 to 4 s	> 4 s
INP (Interaction to Next Paint)	How long the page takes to respond when you tap	≤ 200 ms	200 to 500 ms	> 500 ms
CLS (Cumulative Layout Shift)	How much content moves while loading	≤ 0.1	0.1 to 0.25	> 0.25

To pass you need green on all three at the 75th percentile of real visits. Meaning: it is not enough that it runs well on your machine with fibre. It has to run well for three out of four visitors, many of them on mobile data.

What do you measure with, when tools disagree?

This is the most common confusion, and worth understanding because it changes decisions.

Data type	What it is	Tools	What it is for
Field	What real users experienced over the last 28 days	Search Console, CrUX, PageSpeed Insights (top)	This is what counts for Google
Lab	A simulation under controlled conditions	Lighthouse, PageSpeed Insights (bottom)	Diagnosis, not grading

Lighthouse gives you a 0 to 100 score that is not what Google uses to rank. I have seen teams chase a Lighthouse 100 for weeks while the field data stayed red. Look at Search Console first: it is real field data, grouped by page type.

What should you fix first?

In the order that usually gives the most result per unit of effort:

1. Images. The number one cause of poor LCP. Modern format such as WebP or AVIF, file size matched to display size, and the main image loaded with priority rather than lazily.
2. Reserve space for anything that loads later. If an image, banner or ad appears and pushes text down, that is CLS. Fixed by declaring width and height or a fixed aspect ratio.
3. Fonts. A typeface that loads late produces a text jump. font-display: swap plus preloading the main font resolves it almost always.
4. Third-party JavaScript. Pixels, chats, maps, pop-ups. The main cause of poor INP, because they block the thread that should respond to the user’s tap. Loading them later, or not at all, is the only real fix.
5. Hosting and caching. If the server is slow to respond, everything else arrives late. In Colombia it matters where the server sits and whether there is a delivery network in front.

What else is in technical SEO besides speed?

Core Web Vitals are the visible part. Underneath are things that, if broken, keep you out of the race:

• Indexing. That Google can get in, understands what to index, and does not get lost in duplicates or infinite filter pages.
• Canonicals. That each piece of content has one official address.
• Structured data. The right schema for each page type.
• HTTPS and security headers. Today this is the floor, not a differentiator.
• Rendering. If content only appears with JavaScript, many crawlers never see it. AI crawlers in particular do not execute JavaScript.

That last point became far more important once search stopped being only Google.

How do you run a technical audit without getting lost?

My order, from highest to lowest impact:

1. Can Google get in and understand the site? robots.txt, sitemap, indexing status in Search Console.
2. How many valuable pages are actually indexed? There are almost always surprises here.
3. How are Core Web Vitals in field data, separating mobile and desktop?
4. Is there duplicate content or self-cannibalisation?
5. Is structured data correct and error-free?
6. Does the content appear without JavaScript?

That gives you 80% of the diagnosis. The rest is tuning.

Does this actually move rankings?

Honestly: less than agencies sell and more than developers believe.

Google has said page experience is a tiebreaker, not a replacement for relevance. A slow page with the best answer can beat a fast page with a mediocre one. But between two similar answers, the one that does not make you wait wins.

Where the impact is genuinely large and measurable is conversion. Every second of waiting is paid for in people who leave, and you can see that in your own analytics without having to take anyone’s word for it.

Frequently asked questions

What are Core Web Vitals? Three metrics Google uses to measure page loading experience: LCP, how long the largest element takes to paint; INP, how quickly the page responds to an interaction; and CLS, how much content shifts while loading.

Is FID still used? No. Google replaced it with INP in March 2024. Any guide presenting FID as a current metric is out of date.

What is a good value for each metric? LCP up to 2.5 seconds, INP up to 200 milliseconds and CLS up to 0.1, measured at the 75th percentile of real visits.

Why does PageSpeed Insights give me two different results? Because they are two different things. The top section is real user data from the last 28 days, which is what Google uses. The bottom is a lab simulation, useful for diagnosis. They may not match, and that is not an error.

Is it worth reaching 100 in Lighthouse? It is useful for diagnosis, not for ranking. That score is not what Google evaluates. Better to have 75 in Lighthouse and green field data than the reverse.

What usually damages INP most? Third-party JavaScript: chats, ad pixels, pop-ups and embedded maps. They block the browser exactly when the user taps something.

How long before the data updates after a fix? Up to 28 days, because it is a rolling window of real data. Do not expect to see the change the next day.

Does this change if my traffic is almost all mobile? It changes quite a bit, and in Colombia it almost always is. Measure on mobile and on a constrained connection, not on the office wifi.