Core Web Vitals: loading, responsiveness and stability
Core Web Vitals are three user-experience metrics: LCP measures the appearance of the largest visible content, INP measures interaction responsiveness and CLS measures visual stability. They help describe real visitor experience rather than serving as a general SEO score.
Also known as: Core Web Vitals · CWV
At a glance
- LCP, INP and CLS measure different problems.
- Real-user data and laboratory tests provide different kinds of evidence.
- Good scores do not guarantee rankings, traffic or sales.
On this page

Which experiences do the metrics describe?
A page can remain blank too long, respond slowly to a menu tap or move text while someone reads. Core Web Vitals distinguishes these problems because their causes and remedies can differ.
Page experience matters to SEO, but these metrics do not establish whether information is true, relevant or better than another answer. A fast empty page is not a useful guide.
What are the current thresholds?
| Metric | Experience measured | Upper boundary of the good range |
|---|---|---|
| LCP | Largest visible content element appearing | 2.5 seconds |
| INP | Responsiveness to clicks, taps and keyboard interactions | 200 milliseconds |
| CLS | Unexpected layout shifts | 0.1 |
web.dev's guidance applies assessment at the 75th percentile, considering device groups. CLS is unitless rather than a time value.
A good LCP can coexist with poor responsiveness, and a responsive menu can coexist with unstable content. Read the metrics together without treating them as interchangeable.
What does the 75th percentile mean?
Imagine sorting a hundred comparable page-load LCP observations. The 75th percentile is approximately the value at or below which three quarters of those observations fall. It is neither the fastest test nor simply the average.
A developer's powerful laptop and reliable connection may not represent the audience. Different devices, networks and entry pages produce different experiences. LCP guidance emphasizes examining mobile and desktop separately.
Compare equivalent measurement scopes so a change in audience composition is not mistaken for a technical improvement.
Example: two problems on a course page
On one fictional course page, the hero image appears quickly but the date picker takes a long time to respond. Compressing the image again is not the first conclusion supported by that observation; investigate the interaction delay.
On another version, the date picker responds promptly but a late recommendation banner pushes the registration button away. That is a visual-stability issue.
Both might be described vaguely as a “slow page.” Naming the actual experience helps avoid arbitrary performance work. These are illustrations, not measured Searchboar results.
How do field and laboratory data differ?
A laboratory test runs under controlled conditions and is useful for reproducible diagnosis. Field data comes from actual visits and can represent a broader range of devices and circumstances. Their numbers need not match exactly.
INP considers interactions during the visit, not just initial loading. A test that never uses the menu or form cannot establish their responsiveness.
A new or low-traffic page may lack sufficient real-user data. Missing data is neither a pass nor a failure.
First Contentful Paint, or FCP, is a useful loading diagnostic, but it is not a Core Web Vital. It records when the first text or other qualifying content appears. A fast FCP followed by a slow LCP can mean that a small element appears promptly while the main visible content still takes too long; the earlier paint does not cancel the later delay.
How should a business use the findings?
Connect the metric to important visitor tasks. Investigate loading when the explanation appears late, responsiveness when an action receives delayed feedback, and stability when content moves unexpectedly.
After a fix, compare the same measurement scope. Improved experience can be valuable while search performance and business outcomes still require separate observation. The metrics refine diagnosis; they do not prove the website's overall success.