What is Largest Contentful Paint?

Largest Contentful Paint reports a loading milestone: when a large eligible element becomes visible in the viewport. It uses navigation as its starting point. The reported element can be text, an image or video content; it is not necessarily the element a designer labels as the page’s most important feature.

A booking form can still need further interaction after its large heading is visible. Keep “the main content appeared” separate from “the customer finished the booking.” This distinction prevents a loading metric from becoming an unsupported conversion claim.

Which elements can be candidates?

The LCP reference includes images, SVG image elements, eligible video content, URL-based background images and block-level text. For video, the poster or first frame can supply the measurement, whichever is earlier. A large gradient is not an image candidate simply because it fills the screen.

Visibility and content heuristics matter. Hidden, placeholder or full-viewport background-like elements may be excluded, and portions outside the viewport do not determine the visible size. Check the actual reported candidate; do not assume every hero asset or small icon must be the LCP element.

LCP rating boundaries

The current good boundary is 2.5 seconds or less. For field assessment, use the 75th percentile, segmented by mobile and desktop. The reference defines poor as greater than 4.0 seconds, leaving the middle band above 2.5 through 4.0 seconds.

Published LCP thresholds, in seconds
RatingLCP value
Good≤2.5
Needs improvement>2.5 and ≤4.0
Poor>4.0

A single lab result can be compared with these bands, but it does not establish the field percentile for visitors. Record the environment and reported element alongside its duration.

Where loading time is spent

The LCP optimization guide separates time to first byte, resource load delay, resource load duration and element render delay. The image may finish downloading while the page still waits for styles, script execution or content visibility. Compressing that image alone need not remove the later dependency.

This is why a fixed “use WebP” prescription is incomplete. Inspect when the relevant request is discovered, how it loads and what prevents its display. For a text candidate, also consider the conditions that make its text available to paint.

What one published business experiment showed

A Vodafone case study published on web.dev in March 2021 reports an A/B test of a landing page. The optimized variant had 31% better field LCP and 8% more total sales than the baseline. Paid-media traffic was split equally, and the two versions were visually and functionally identical.

The optimization combined changes to widget rendering, critical HTML and images. This is a reported result for that experiment, not proof that LCP alone caused the sales difference or a forecast for another business. The source redacts absolute business figures. This page cites the third-party report without claiming an independent reproduction.

Choose a change that fits the dependency

If an LCP image is discovered late, check whether its actual source is available in the initial response and whether lazy loading delays the request. If the resource is already ready, investigate the work preventing its render. The official guide cautions against lazy-loading the LCP image.

Treat a priority hint, stylesheet reduction, delivery change or script change as a testable intervention. Retest both the reported element and the reader’s task. Avoid adding high priority to unrelated assets or changing loading order without checking the interface.

Check field and lab evidence

PageSpeed Insights presents historical CrUX data separately from its lab diagnostics. It may use origin data when the exact URL has too few samples, or show no field data when the origin is also insufficient. Inspect the scope label before calling the result a measurement of your page.

Use a controlled load to inspect the candidate and loading sequence, then compare it with available field evidence. A result from one cached desktop test cannot describe every mobile visitor. The measurement guidance explains why local and real-user results may differ.

LCP and First Contentful Paint

First Contentful Paint describes the appearance of initial content; LCP is more selective about large content. A small header can appear before a larger product photo. A fast first content paint therefore does not establish a fast LCP. Neither timing tells you when every application function is ready.

FCP and TTFB are supporting loading diagnostics, not extra Core Web Vitals. Use them to investigate where the wait occurs rather than treating their labels as interchangeable with LCP.

Keep comparisons meaningful

  • Record the URL, viewport and candidate element.
  • Note cache, network and device conditions for local tests.
  • Keep URL field data distinct from origin aggregates.
  • Check the changed dependency rather than just a headline score.
  • Retest after changing templates or important third-party code.

Avoid invented outcomes

Do not attach a universal bounce-rate, sales or ranking change to an LCP improvement. Google uses Core Web Vitals in ranking systems, but its page-experience guidance does not guarantee top rankings. A published case from another business is not a forecast for this page.

LCP includes navigation and response delays. Identify the server-response portion and the remaining loading sequence before attributing a slow LCP to a single dependency.

Check the business page people need

Check a booking or quote page separately from the homepage. Record the page, viewport and observed candidate before changing an image; a template’s “hero” label does not identify the measured LCP element. Retest whether the page’s actual task still works after changing delivery or rendering.

Additional measurement questions

Can the reported LCP element change during loading?

Yes. The browser can first report a visible text block and later a larger eligible image after it renders. That makes the candidate sequence useful diagnostic evidence. Inspect the last relevant recorded candidate rather than assuming that the first element reported is the final loading milestone for that navigation.

Is a hero image always the LCP element?

No. The candidate depends on eligible content that is actually visible and its size, not a “hero” class or a designer’s intention. A text block or eligible video can be reported instead. Check the observed candidate for the tested viewport before changing an image that might not control this measurement.

See Core Web Vitals for the shared field assessment, render-blocking resources for paint dependencies and server-side rendering for initial HTML delivery.

Sources and editorial note

Primary references are linked beside the relevant explanations and were checked on October 12, 2026. AI assisted this revision. The Vodafone figures are attributed to the published third-party case; other examples are illustrative. No first-party measurement, independent performance experiment, private Search Console evidence or named human approval is claimed.