What is INP?

Interaction to Next Paint measures responsiveness across qualifying interactions during a page visit. Its latency includes waiting to start event handling, running that handling and waiting for the next frame. It concerns prompt visual feedback, not the completion time of every later network request or operation started by the interaction.

When someone opens a menu or submits a form, there are two questions: did the interface respond, and did the full operation succeed? A prompt response helps the first question; it cannot answer the second. This distinction matters when interpreting an interaction trace.

Responsiveness and search outcomes

INP is part of the current Core Web Vitals set. Google says those metrics are used by ranking systems, while good page-experience results do not guarantee top rankings. No fixed conversion loss, bounce increase or ranking lift follows from an INP band.

Use the score to identify delayed feedback and investigate affected tasks. A filter that seems unresponsive deserves attention because the visitor cannot tell whether their input was accepted. Measure any change in business outcomes separately.

Qualifying interactions and visit-level values

The INP definition observes clicks, taps and keyboard interactions. Scrolling and hovering are not themselves qualifying INP interactions, though a related gesture may include a click or keypress. An interaction can involve several event callbacks; the measurement concerns that logical gesture.

The visit-level value generally reflects the slowest interaction, with specified outlier handling on pages with many interactions. It is not correct to describe every visit as a simple 98th-percentile calculation. Separately, the field assessment uses the 75th percentile of page loads, segmented by mobile and desktop.

A visit without qualifying input may have no INP value. That is not a zero-latency result. The reference includes non-interacting users and an unscripted headless browser among reasons for a missing observation.

INP rating boundaries

The published boundaries include exactly 200 milliseconds in the good band and exactly 500 milliseconds in needs improvement. These are latency thresholds, not measured outcomes for a hypothetical customer or a promise that every interaction will feel immediate.

Published INP thresholds, in milliseconds
RatingINP value
Good≤200
Needs improvement>200 and ≤500
Poor>500

A field percentile and a single interaction trace are different observations. Label which one you have before applying the bands. Do not call a value in the middle band poor or assume a self-selected safety margin guarantees the field assessment.

An illustrative delayed filter

Imagine a product filter whose interface does not acknowledge a tap until unrelated work completes. This is an illustrative scenario, not a customer test. First reproduce that gesture, then inspect whether the wait happens before its event callbacks, during their work or before the next frame is presented.

The INP optimization guide uses those input-delay, processing-duration and presentation-delay subparts to locate the cause. A later product request can still take time even when the interface provides prompt feedback; INP does not measure every eventual effect.

INP addresses input responsiveness; LCP concerns loading and CLS concerns unexpected layout movement. They use different measurements. The current Core Web Vitals set does not collapse them into one equal-weight score.

First Input Delay measured only the initial input-delay portion of the first interaction. INP considers qualifying interactions throughout the visit and includes processing and presentation. That broader scope does not support a universal claim that one numerical threshold is “harder” for every site.

INP replaced First Input Delay as a Core Web Vital in March 2024. The official launch announcement documents the transition. Use current metric definitions when reading an older report that still names FID.

Investigate before changing code

  1. Identify the page and gesture that provide slow feedback.
  2. Record available field context and the tested device.
  3. Reproduce the interaction during relevant loading or runtime conditions.
  4. Locate the delayed subpart before choosing a change.
  5. Retest the full task, then monitor available field evidence.

The official investigation guidance starts from field observations where available, then uses lab work to diagnose. Reducing unnecessary handler or rendering work can be relevant; a blanket script deferral or DOM-node limit is not a substitute for this diagnosis.

Do not confuse no-input audits with INP

A standard navigation-only Lighthouse run has no user input and cannot measure INP. Its Total Blocking Time result is a useful proxy for possible problems, not an observed interaction latency. Lab INP testing is possible when meaningful interactions are actually exercised.

PageSpeed Insights separates real-user and lab data, and may use origin-level field data when the page lacks samples. Check those labels. Neither a passing local gesture nor a missing field result proves that all visitors receive good responsiveness.

Additional measurement questions

Does INP include the entire wait for a server operation?

Not necessarily. INP concerns the next frame following a qualifying interaction, not every later network fetch or asynchronous update it starts. An interface can provide prompt feedback while a server operation is still running. Check both the interaction latency and eventual task completion rather than presenting one as evidence of the other.

Can INP be investigated without field data?

Yes. You can exercise representative user flows in a controlled environment and inspect slow interactions there. That provides diagnostic evidence under recorded conditions, not a real-user percentile. Keep field data unavailable when it is absent, and state the tested gesture, device and loading conditions when reporting a local result.

Read Core Web Vitals for the shared assessment, LCP for loading and JavaScript SEO for crawler content access. Crawler access and interaction speed need separate evidence.

Sources and editorial note

Primary references are linked beside the relevant explanations and were checked on October 12, 2026. AI assisted this revision. Examples are illustrative. No customer measurement, independent performance experiment, private Search Console evidence or named human approval is claimed.