Technical SEO · Supporting checklist

Core Web Vitals Checklist

A Core Web Vitals checklist is a performance review of real-user loading, responsiveness and visual stability. Read field data for the correct URL or origin and device, then diagnose LCP, INP and CLS on representative templates. The good thresholds are 2.5 seconds, 200 milliseconds and 0.1 respectively at the 75th percentile. Keep immediate lab checks separate from later field confirmation.

Free to useNo account neededSaved in your browser
Three page models representing content loading, a button interaction and a stable layout, connected to a group of phones.
Review loading, interaction responsiveness and visual stability using real-user field data.
Your working checklist

Work through the checklist

Click a square to mark a task as Pass. Click again to undo. Use the result menu for a problem, a blocker or a task that does not apply. You do not need to enter notes.

15 checks

Read Field Data Before Choosing a Fix

Field data describes collected real-user experiences. Identify the population, device, period and missing metrics before deciding which template or metric needs attention.

Identify the field data and its exact scope

Know whose experience a number describes.

Use this for: The representative public pages selected for this review.

Run an important public page through PageSpeed Insights and look first at the real-user section. Check whether the report describes the URL itself or the whole origin, meaning the shared protocol and hostname. Keep mobile and desktop results separate.

Note the shown collection period and available metrics. An origin result cannot establish that every page template performs the same way. If field data is unavailable, keep that status explicit and use lab tests to investigate; missing data is not a good result.

Example

Hypothetical example: A new product URL shows origin-level field data. The team uses it as site context and tests the product template separately instead of calling that URL passed.

  1. Open the real-user section for the selected URL.
  2. Identify URL versus origin, device and collection period.
  3. Keep unavailable metrics distinct from measured results.
More help and optional notesHigh · SEO analyst
PassEach field result is labeled with its actual population, device and period.
FailOrigin data, mobile data or missing data is presented as a measured result for a different population.

If it fails: Correct the scope labels before prioritizing fixes.

Retest: Recheck the scope whenever another URL or report is used.

Not applicable: No pages or features in the stated scope require this check.

Optional: the URL or template, what you found and the next action. Keep confidential data out of shared exports.

No result recorded yet.

Changing this result updates your checkmark and progress immediately. Notes are optional.

Classify LCP, INP and CLS separately

Find the metric that actually needs work.

Use this for: The representative public pages selected for this review.

Compare the reported 75th-percentile values with the good thresholds: LCP at or below 2.5 seconds, INP at or below 200 milliseconds, and CLS at or below 0.1. The 75th percentile means that 75% of measured experiences are at that value or better; it is not the average.

Keep the metric values alongside the report’s assessment. A fast LCP does not cancel a poor INP or CLS. Follow the displayed insufficient-data rules rather than inventing a missing score. Completing this measurement check does not mean the page has passed Core Web Vitals.

Example

Hypothetical example: A mobile report has LCP of 2.2 seconds, INP of 310 milliseconds and CLS of 0.04. Responsiveness needs investigation even though loading and stability are good.

  1. Read the individual field values and units.
  2. Compare each available metric with its threshold.
  3. Prioritize the failing metric instead of averaging the three.
More help and optional notesHigh · SEO analyst
PassThe available metrics are classified correctly without substituting an average or lab score.
FailA good metric hides a failing one or the wrong units change the classification.

If it fails: Correct the classification and choose the affected metric for diagnosis.

Retest: Repeat against the next comparable field period.

Not applicable: No pages or features in the stated scope require this check.

Optional: the URL or template, what you found and the next action. Keep confidential data out of shared exports.

No result recorded yet.

Changing this result updates your checkmark and progress immediately. Notes are optional.

Map affected URL groups to real templates

Fix shared causes without treating one example as the whole site.

Use this for: The representative public pages selected for this review.

In Search Console’s Core Web Vitals report, open the relevant mobile or desktop issue and inspect its example URL groups. Match several examples to actual templates such as articles, products and categories. Add important traffic or business journeys that the available sample might not represent.

Prioritize a shared defect that affects many important pages, then confirm it on more than one example. Search Console groups and individual PageSpeed Insights results can differ; the group is not a separate lab test of every listed URL.

Example

Hypothetical example: Several poor-LCP examples use the same product gallery. The team checks two product pages before assigning the shared gallery fix.

  1. Open the device-specific issue and example groups.
  2. Map examples to templates and important journeys.
  3. Confirm the shared symptom on representative pages.
More help and optional notesHigh · SEO analyst and developer
PassThe repair queue identifies affected templates and the scope supported by the report.
FailOne example is treated as proof about every route or all devices.

If it fails: Sample the missing template and refine the affected scope.

Retest: Revisit the group and representative URLs after the shared fix.

Not applicable: No pages or features in the stated scope require this check.

Optional: the URL or template, what you found and the next action. Keep confidential data out of shared exports.

No result recorded yet.

Changing this result updates your checkmark and progress immediately. Notes are optional.

Create a repeatable diagnostic test

Use the lab to explain a field symptom.

Use this for: The representative public pages selected for this review.

Test the selected URL in PageSpeed Insights or Chrome DevTools. Keep the device mode, throttling, cache state and relevant page state consistent when comparing runs. Repeat the test if a result is an obvious outlier rather than choosing the best score.

Open the detailed trace or diagnostic finding for the failing metric. Lab runs simulate a particular visit; field data contains different devices, networks and longer interactions. Treat a disagreement as something to investigate, not a reason to discard the worse result.

Example

Hypothetical example: A quick lab load looks stable, but field CLS is poor. A longer recording through the page’s ad-loading and scrolling states reveals a later shift.

  1. Choose the failing template and a consistent test setup.
  2. Repeat the relevant load or interaction.
  3. Use the trace to look for a cause that matches the field symptom.
More help and optional notesHigh · Developer and SEO analyst
PassThe diagnostic comparison uses a repeatable setup and preserves field/lab distinctions.
FailThe team cherry-picks a score or compares incompatible test conditions.

If it fails: Repeat with the same conditions and inspect the actual trace.

Retest: Use that setup again after the repair.

Not applicable: No pages or features in the stated scope require this check.

Optional: the URL or template, what you found and the next action. Keep confidential data out of shared exports.

No result recorded yet.

Changing this result updates your checkmark and progress immediately. Notes are optional.

LCP Checklist: Find the Main-Content Delay

LCP checks identify the largest eligible visible content and explain why it appeared late. Diagnose the timing stage before assigning a resource fix.

Find the actual LCP element

Optimize what became the largest visible content.

Use this for: The representative public pages selected for this review.

Open the LCP detail in the performance report or trace and identify the element for the tested viewport. It may be an image or a text block. Confirm its location on the page rather than assuming every template’s hero image is the LCP element.

Compare mobile and desktop where both need work. If a banner or layout change produces a different LCP element, investigate that actual state. Do not remove useful content merely to make a metric look smaller.

Example

Hypothetical example: On desktop the product photo is LCP, but on mobile a large title block is LCP. The mobile investigation focuses on its font and render path.

  1. Open the reported LCP element detail.
  2. Match it to the visible element and viewport.
  3. Check whether important page states use a different element.
More help and optional notesHigh · SEO lead and developer
PassThe identified element matches the tested page and viewport.
FailThe fix targets a guessed hero that is not the measured LCP element.

If it fails: Inspect the element attribution before selecting a repair.

Retest: Confirm the element again after a layout change.

Not applicable: No pages or features in the stated scope require this check.

Optional: the URL or template, what you found and the next action. Keep confidential data out of shared exports.

No result recorded yet.

Changing this result updates your checkmark and progress immediately. Notes are optional.

Locate the delay inside LCP

Separate waiting for the response, discovery, download and display.

Use this for: The representative public pages selected for this review.

For an image LCP, inspect time to first byte, resource load delay, resource load duration and element render delay. In plain terms, these cover waiting for the page response, discovering the image, downloading it and finally displaying it. Use the trace’s breakdown rather than guessing from file size.

For text LCP, investigate the response and rendering path, including blocking CSS or fonts where applicable. Give the developer the largest supported delay. Compressing an image may not move LCP if scripts still keep the finished image from being painted.

Example

Hypothetical example: A small banner image downloads quickly but is inserted only after a large script runs. Earlier discovery is the first experiment, rather than further image compression.

  1. Inspect the LCP timing breakdown or relevant request trace.
  2. Identify the dominant waiting or rendering stage.
  3. Choose one repair that targets that stage.
More help and optional notesHigh · SEO lead and developer
PassThe proposed repair is tied to an observed part of the LCP delay.
FailEvery LCP problem is assigned to image size without checking the timing.

If it fails: Inspect the missing timing stage and revise the diagnosis.

Retest: Compare the same breakdown after the change.

Not applicable: No pages or features in the stated scope require this check.

Optional: the URL or template, what you found and the next action. Keep confidential data out of shared exports.

No result recorded yet.

Changing this result updates your checkmark and progress immediately. Notes are optional.

Retest the LCP cause after a targeted repair

Check whether the visitor sees the main content sooner.

Use this for: The representative public pages selected for this review.

Ask the responsible developer to apply the selected fix, such as improving the response, exposing the critical image earlier or removing a render delay. Use the page-speed checklist for the detailed resource work.

Rerun the same template and conditions, then compare both total LCP and the timing stage you intended to improve. Check that the image quality, useful content and other metrics remain acceptable. An improved download alone is insufficient if total LCP did not improve.

Example

Hypothetical example: A change cuts image download time but the text-and-image reveal still waits for a script. The task stays open until that remaining render delay is addressed.

  1. Implement the cause-specific repair.
  2. Compare total LCP and the targeted timing stage.
  3. Check content, image quality and the other metrics.
More help and optional notesHigh · SEO lead and developer
PassThe matched test shows the intended LCP improvement without a new functional or visual regression.
FailOnly a subpart improves while the same bottleneck still delays visible content.

If it fails: Revisit the remaining delay instead of stacking unrelated optimizations.

Retest: Confirm the effect later in comparable field data.

Not applicable: No pages or features in the stated scope require this check.

Optional: the URL or template, what you found and the next action. Keep confidential data out of shared exports.

No result recorded yet.

Changing this result updates your checkmark and progress immediately. Notes are optional.

INP Checklist: Trace Slow Interactions

INP checks follow the visitor’s clicks, taps and typing through the page’s response. A fast initial load can still have a slow interactive journey.

Reproduce slow taps, clicks and typing

Measure responsiveness during the actions visitors take.

Use this for: The representative public pages selected for this review.

Choose interactions from the affected template: opening the menu, changing a product option, applying a filter or typing into search. Record them in the Performance panel, including early actions while the page is still busy and later actions after it has loaded.

INP concerns click, tap and keyboard interactions across the visit. A navigation-only Lighthouse run does not directly establish field INP; Total Blocking Time can help diagnose main-thread work, but it is not an interchangeable INP result.

Example

Hypothetical example: A category page loads quickly but stalls when a filter is selected. The team records that action instead of repeatedly testing only the initial load.

  1. Choose important interactions on the failing template.
  2. Record early and later interactions under consistent conditions.
  3. Identify a reproducible delayed visual response.
More help and optional notesHigh · SEO lead and developer
PassThe slow interaction is reproducible, or the missing reproduction is explicitly left unresolved.
FailA load-only score is used as proof that interactive responsiveness is good.

If it fails: Capture the actual interaction or add suitable existing real-user attribution.

Retest: Repeat the same interaction after the repair.

Not applicable: No pages or features in the stated scope require this check.

Optional: the URL or template, what you found and the next action. Keep confidential data out of shared exports.

No result recorded yet.

Changing this result updates your checkmark and progress immediately. Notes are optional.

Find which part of the interaction is slow

Assign the fix to the work that blocks feedback.

Use this for: The representative public pages selected for this review.

Ask the developer to inspect the slow interaction’s input delay, event-handler processing and presentation delay. These mean waiting to start handling the action, running its code, and waiting to display the next frame. Check whether a third-party script or a large layout update overlaps the action.

Choose a targeted change: reduce competing work, shorten or divide the handler, or simplify the rendering work. Keep immediate visual feedback meaningful, such as showing a selected filter or a loading state, while the required operation continues.

Example

Hypothetical example: A filter click updates hundreds of cards and recalculates their layout before painting the selected state. The developer narrows the update and separates nonurgent work.

  1. Locate the slow interaction in the trace.
  2. Identify waiting, handler work or presentation work.
  3. Choose the responsible component and targeted repair.
More help and optional notesHigh · SEO lead and developer
PassThe fix addresses a measured interaction stage and responsible component.
FailThe team removes unrelated files without identifying the slow action’s work.

If it fails: Trace the interaction again and isolate the blocking work.

Retest: Compare the same interaction timing after the change.

Not applicable: No pages or features in the stated scope require this check.

Optional: the URL or template, what you found and the next action. Keep confidential data out of shared exports.

No result recorded yet.

Changing this result updates your checkmark and progress immediately. Notes are optional.

Retest feedback and completion after the fix

Improve responsiveness without breaking the action.

Use this for: The representative public pages selected for this review.

Repeat the slow action with the same data and test setup. Check that the page displays feedback promptly and still completes the intended operation. Test rapid repeated input and an error state where relevant; a disabled or incomplete feature is not a performance success.

Compare the local interaction trace, then plan a field review for the affected device and template. Keep privacy-sensitive form values out of shared recordings or optional notes.

Example

Hypothetical example: A search field feels faster after its work is deferred, but stale results appear after quick typing. The fix needs request-order handling before it is complete.

  1. Repeat the original action and compare its trace.
  2. Test repeated input, correct results and relevant errors.
  3. Keep field verification separate from the local result.
More help and optional notesHigh · SEO lead and developer
PassThe interaction is more responsive in the matched test and its result remains correct.
FailFeedback improves by dropping work, returning stale content or breaking the journey.

If it fails: Restore correct behavior and remove the measured bottleneck safely.

Retest: Verify the corrected journey and later field trend.

Not applicable: No pages or features in the stated scope require this check.

Optional: the URL or template, what you found and the next action. Keep confidential data out of shared exports.

No result recorded yet.

Changing this result updates your checkmark and progress immediately. Notes are optional.

CLS Checklist: Find Unexpected Movement

CLS checks trace unexpected layout shifts through the visit. Reserve space and repair the component that causes movement, rather than only inspecting the element that moved.

Record shifts beyond the first screen

Find unexpected movement during the visit.

Use this for: The representative public pages selected for this review.

Record the page loading, then scroll and use the relevant controls while late content, embeds or promotions appear. Inspect the layout-shift events and affected elements. A load-only test can miss shifts that occur later in a visit.

Look for the element that caused the movement, not only the text or button that moved. CLS is not a simple sum of every visual movement; the metric has defined session windows and excludes some shifts associated with recent input. Do not hide a problem by relying on an assumption about that exclusion.

Example

Hypothetical example: A review widget expands above a reading position after the initial load. The article text moves, but the widget’s unreserved space is the cause.

  1. Record loading plus realistic scrolling and actions.
  2. Inspect the largest unexpected shift events.
  3. Identify the component that changed the layout.
More help and optional notesHigh · SEO lead and developer
PassThe relevant unexpected shifts have a reproducible trigger and identified cause.
FailA stable first-screen load is treated as proof that the full visit is stable.

If it fails: Extend the recording through the delayed content or journey state.

Retest: Repeat the complete sequence after the repair.

Not applicable: No pages or features in the stated scope require this check.

Optional: the URL or template, what you found and the next action. Keep confidential data out of shared exports.

No result recorded yet.

Changing this result updates your checkmark and progress immediately. Notes are optional.

Reserve space for images, ads and embeds

Keep late arrivals from pushing existing content away.

Use this for: The representative public pages selected for this review.

For the component causing a shift, inspect whether the page reserves its expected dimensions before it loads. Ask the developer to provide image dimensions or an appropriate aspect ratio and a planned container size for embeds and ad slots.

Test responsive variants and empty or failed responses too. A slot that initially reserves space but collapses later can still move content. Choose space based on the actual component and layout rather than applying the same fixed height everywhere.

Example

Hypothetical example: A map embed has no initial height, so the contact details jump down when it loads. A responsive reserved container stabilizes the page.

  1. Identify the shifting media or slot.
  2. Reserve appropriate space before it loads.
  3. Test loaded, empty and failed states at relevant widths.
More help and optional notesHigh · SEO lead and developer
PassThe component’s arrival or fallback no longer causes the reproduced unexpected movement.
FailSpace appears late or collapses unexpectedly in a common state.

If it fails: Correct the dimensions and fallback behavior.

Retest: Record the same loading sequence again.

Not applicable: No pages or features in the stated scope require this check.

Optional: the URL or template, what you found and the next action. Keep confidential data out of shared exports.

No result recorded yet.

Changing this result updates your checkmark and progress immediately. Notes are optional.

Check font swaps and inserted interface content

Catch shifts left after media dimensions are fixed.

Use this for: The representative public pages selected for this review.

Watch headings and paragraphs as fonts load, and test banners, validation messages and dynamically inserted sections. If a font swap changes wrapping, ask the developer to compare fallback metrics and the chosen font-loading strategy.

For inserted content, reserve a stable area where practical or place the change where it does not unexpectedly displace existing reading content. Keep required messages visible. Do not remove an error notice or consent control solely to reduce the metric.

Example

Hypothetical example: A late brand font makes the title wrap to three lines and shifts the first paragraph. A better-matched fallback reduces the movement without hiding the heading.

  1. Record font loading and dynamic interface states.
  2. Identify changed wrapping or newly inserted space.
  3. Repair the fallback or layout while retaining useful messages.
More help and optional notesHigh · SEO lead and developer
PassThe reproduced font or dynamic-content shift is resolved without hiding necessary information.
FailThe layout still jumps or essential messages were removed to suppress movement.

If it fails: Adjust fallback metrics or the component’s placement and reserved space.

Retest: Test fresh font loading and the original dynamic trigger.

Not applicable: No pages or features in the stated scope require this check.

Optional: the URL or template, what you found and the next action. Keep confidential data out of shared exports.

No result recorded yet.

Changing this result updates your checkmark and progress immediately. Notes are optional.

Verify the Repair and Watch for Regressions

Verification combines a repeatable immediate test with a later review of the relevant field population. Keep unresolved field outcomes explicit.

Separate immediate retests from field confirmation

Avoid closing a real-user issue on one fresh lab result.

Use this for: The representative public pages selected for this review.

After the fix is live, retest the affected public templates immediately to catch regressions. Keep the release date alongside the change. Then review the same field population and device as new visits enter its reporting window.

PageSpeed Insights field data represents a trailing 28-day period, so early results can contain both old and new experiences. Search Console’s issue validation uses a 28-day monitoring period and does not request reindexing. Keep a successful lab repair and pending field confirmation as separate states.

Example

Hypothetical example: A release on the 10th improves the repeatable mobile trace, while the field report on the 11th still contains many pre-release visits. The team keeps monitoring the same cohort.

  1. Check the live repair with the original diagnostic test.
  2. Record the release date and comparable field population.
  3. Review the later field result or applicable validation status.
More help and optional notesHigh · SEO analyst and developer
PassThe live repair is checked and field confirmation is either supported by comparable data or clearly pending.
FailA same-day lab score is labeled a confirmed field pass.

If it fails: Restore the pending field status and compare the correct window.

Retest: Review until the relevant post-change field evidence is available.

Not applicable: No pages or features in the stated scope require this check.

Optional: the URL or template, what you found and the next action. Keep confidential data out of shared exports.

No result recorded yet.

Changing this result updates your checkmark and progress immediately. Notes are optional.

Set a focused performance regression check

Keep the repaired template from silently getting worse.

Use this for: The representative public pages selected for this review.

Keep a short set of representative URLs and actions for releases that change the affected templates. Use comparable automated or manual lab tests for quick feedback, then review field trends for the devices and journeys that matter.

If values worsen, check recent image, font, plugin, tag or layout changes and any change in visitor mix. Prefer a reproducible regression to a one-off score fluctuation. Agree who owns the repair and the next check; this checklist does not start monitoring your site automatically.

Example

Hypothetical example: A new chat widget overlaps product-filter interactions. The saved filter journey makes the regression easier to reproduce and assign.

  1. Retain the relevant URL and interaction sample.
  2. Run it after meaningful template or dependency changes.
  3. Investigate sustained field changes or reproducible lab regressions.
More help and optional notesMedium · SEO lead and developer
PassThe repaired workflow has a named owner and a practical repeatable check.
FailThe issue is closed with no way to detect a recurring template defect.

If it fails: Add the affected URL and action to the release review.

Retest: Revisit the sample when templates or important dependencies change.

Not applicable: No pages or features in the stated scope require this check.

Optional: the URL or template, what you found and the next action. Keep confidential data out of shared exports.

No result recorded yet.

Changing this result updates your checkmark and progress immediately. Notes are optional.

Reset this checklist?

This removes saved statuses, evidence and result dates for the checks on this page. Other checklists keep their records. Export a CSV first if you need a copy.

A useful result, every time

Check a task. Choose a result.

This is a manual SEO workflow. It helps you decide what to investigate and keep a usable record. It does not crawl your website or verify your answers automatically.

  1. Choose your scope. Pick the relevant task group and representative pages you are authorized to inspect.
  2. Run the stated check. Follow the steps, then choose your result. Click the square for Pass, or use the result menu for another status. You do not need to type anything.
  3. Assign the next action. Fix a problem when you can, or ask the right person for help. Optional notes can remind you what to do next. Check the result again after a fix.

What your progress means

Review progress counts passes and failures against applicable checks. Failed work still counts as reviewed. Blocked and untested checks remain unfinished. This is not an SEO score.

Untested
You have not checked this yet.
Pass
You checked it and it works as described.
Fail
You checked it and found something to fix.
Blocked
You need access, information or someone’s help.
Not applicable
This task does not apply to your website.

Export CSV downloads the checks currently shown and any optional notes. Print expands the instructions for that view. Saved results stay in this browser; there is no account sync.

Apply the workflow

How Should You Interpret the Core Web Vitals Results?

Use the thresholds, data-availability rules and implementation handoff below to keep your conclusions within what the measurements actually show.

What Are the Good Core Web Vitals Thresholds?

Good Core Web Vitals thresholds are LCP at or below 2.5 seconds, INP at or below 200 milliseconds and CLS at or below 0.1. Compare the 75th percentile for the relevant device and population.

The table classifies individual metric values. Follow the tool’s overall assessment and missing-data rules separately.

Core Web Vitals thresholds for individual 75th-percentile values.
MetricGoodNeeds improvementPoor
LCP: loading≤ 2.5 s> 2.5 s to 4 s> 4 s
INP: responsiveness≤ 200 ms> 200 ms to 500 ms> 500 ms
CLS: visual stability≤ 0.1> 0.1 to 0.25> 0.25

What Does a Useful Core Web Vitals Backlog Look Like?

A useful Core Web Vitals backlog connects a field symptom to a reproducible cause, a specific repair and two verification states: immediate diagnostic testing and later field confirmation.

Use the hypothetical examples below as a format. The checklist’s optional notes and CSV export do not collect performance data or run website tests.

Hypothetical measured symptoms and targeted repairs.
Observed symptomDiagnostic findingTargeted repairVerification
Mobile product LCP needs improvement.Gallery image discovered after app execution.Expose the initial image earlier.Compare LCP trace now; review the same field cohort later.
Category INP is poor.Filtering rebuilds a large layout before feedback.Reduce the update and divide nonurgent work.Repeat filtering; verify correct results and field trend.
Article CLS is poor.A delayed embed inserts unreserved space.Reserve space and test the fallback.Record a full reading journey and review field CLS.

What If PageSpeed Insights Has No Field Data?

No field data means the report cannot provide that measurement for the requested scope. The page may be new, have too little eligible traffic or otherwise lack a reportable sample. It is not a pass or a failure.

Use repeatable lab tests now and any suitable, already-approved real-user monitoring you have. If PSI shows origin data instead, label it as origin context. Do not change privacy controls simply to manufacture a sample.

Can an Assessment Pass When INP Is Missing?

PageSpeed Insights can pass an assessment using good LCP and CLS values when INP has insufficient data. If LCP or CLS is also unavailable, the aggregation cannot be assessed under its documented rules.

Keep the missing INP measurement visible in your review and test important interactions separately. A displayed pass with missing INP is not evidence that every interaction has been measured.

Does a High Lighthouse Score Guarantee Good Core Web Vitals or Rankings?

A high Lighthouse score does not guarantee good field Core Web Vitals or rankings. Lighthouse measures a particular simulated visit, while field data describes collected real-user experiences.

Google uses Core Web Vitals in its ranking systems, but strong scores do not guarantee a top position. Fix the experience problem while continuing to provide relevant, useful content.

Which Checklist Handles the Implementation Work?

Use the diagnosis here to choose the next implementation step, then return to the same metric and population for verification.

A clear process, with honest limits.

Follow the steps, choose the result that matches what you found, and check again after changes. Notes are optional. A completed check does not guarantee indexing, traffic or rankings. How the checklists work.