Core Web Vitals & PageSpeed Monitor – Real User CrUX, History & Lighthouse

Check Core Web Vitals (LCP, INP, CLS, FCP, TTFB) for any list of URLs or domains directly from Google's Chrome UX Report (CrUX) API and CrUX History API (up to 40 weeks of weekly trends from the very first run), with optional PageSpeed Insights (Lighthouse) audits for lab diagnostics. Schedule it with a syncKey to receive regression alerts in Slack or your workflow whenever real-user metrics cross Google's thresholds or degrade.

Why this Actor

• Real Chrome user data first (mode: "field"). Most PageSpeed tools run a slow Lighthouse lab test on every URL and read CrUX only as a side field of PageSpeed Insights — which Google has announced it will remove from the PSI API. This Actor queries the Chrome UX Report API and CrUX History API directly: thousands of URLs in minutes, 28-day rolling p75 + good / needs-improvement / poor distributions, and a 25–40 week historical trend on the very first run.
• Page-level and domain-level data Together. Smaller pages and subpages often lack URL-level CrUX traffic. Each row includes both urlMetrics and originMetrics (queried once per domain per run) with an explicit fieldSource: "url" | "origin" | "none".
• Redirect normalization before CrUX. The CrUX API does not follow redirects (http://wikipedia.org returns 404, while https://www.wikipedia.org/ has full data). This Actor resolves redirects and checks DNS/HTTP status before querying CrUX or PageSpeed Insights.
• Noise-free monitoring alerts. Single-run Lighthouse scores fluctuate by 5–28 points between identical runs. This Actor bases monitoring alerts (syncKey) on stable 28-day CrUX field metrics: it alerts when a page drops from passed to failed, crosses a Google threshold (good → needs-improvement / poor), or worsens by at least minChangePercent.
• Lighthouse diagnosis when you need it (mode: "audit"). Adds Google PageSpeed Insights v5 (Lighthouse 13): 4 category scores (performance, accessibility, bestPractices, seo), lab metrics (lcpMs, cls, tbtMs, fcpMs, speedIndexMs, ttfbMs) and the top 3–5 actionable opportunities with estimated millisecond and byte savings.
• Built-in sitemap discovery and domain report. Pass domains with fromSitemap: true to automatically discover and check up to maxPagesPerDomain URLs from /robots.txt and /sitemap.xml. Every run writes a domain-level roll-up (domainsReport, worstPages, regressions) to OUTPUT.

Who uses it

• SEO & web performance teams: track Core Web Vitals pass rates across key templates, product pages and sitemaps.
• Agencies: automated weekly client reports with 25–40 week CrUX trends and Slack regression digests.
• Lead generation & sales: enrich domain lists with real-user mobile speed (cwvAssessment: "failed", slow LCP/INP) to find sites that need optimization.
• AI agents & MCP workflows: structured JSON records, clear statusReason codes and a run-level OUTPUT summary.

Quick start

{
    "urls": ["https://www.wikipedia.org/", "https://developer.chrome.com/"],
    "mode": "field",
    "devices": ["phone", "desktop"]
}

Goal	Input
Fast real-user check (CrUX + 25-week trend)	{"urls": ["https://www.wikipedia.org/", "https://shopify.com/"], "mode": "field"}
Full Lighthouse + CrUX audit	{"urls": ["https://www.wikipedia.org/"], "mode": "audit", "devices": ["phone", "desktop"]}
Audit a domain from its sitemap	{"domains": ["developer.chrome.com"], "fromSitemap": true, "maxPagesPerDomain": 25, "mode": "field"}
Daily/weekly monitoring, deliver only changes	{"urls": ["https://www.wikipedia.org/"], "syncKey": "prod-sites", "onlyChanges": true}
Slack alerts when Core Web Vitals degrade	{"domains": ["example.com"], "fromSitemap": true, "syncKey": "client-a", "slackWebhookUrl": "https://hooks.slack.com/services/…"}
40-week historical CrUX trend	{"urls": ["https://www.wikipedia.org/"], "historyWeeks": 40}
Strict custom thresholds (e.g. LCP ≤ 2000 ms)	{"urls": ["https://www.wikipedia.org/"], "lcpGoodMs": 2000, "inpGoodMs": 150}
Large Lighthouse audit with your own Google key	{"domains": ["example.com"], "fromSitemap": true, "maxPagesPerDomain": 250, "mode": "audit", "googleApiKey": "AIza…"}

Input

Field	What it does
urls	Page URLs to check. HTTP redirects are followed before querying CrUX and PSI.
domains	Domains or origins (e.g. wikipedia.org). Checks the homepage and origin CrUX; combine with fromSitemap: true to sample pages from sitemaps.
fromSitemap, maxPagesPerDomain	Discover URLs from /robots.txt and /sitemap.xml (including <sitemapindex>), up to maxPagesPerDomain (default 50).
mode	field (default: CrUX + History API, fast) or audit (adds PageSpeed Insights Lighthouse lab audit).
devices	phone, desktop (default both), plus tablet in field mode.
includeHistory, historyWeeks	Include weekly CrUX History trend (default on, 25 weeks; configurable 1–40 weeks).
syncKey, onlyChanges, syncReset	Monitoring across runs in your account's key-value store (cwv-monitor-sync-<key>).
minChangePercent	Minimum percentage p75 worsening (default 10%) to flag a metric as degraded.
lcpGoodMs, lcpPoorMs, inp…, cls…	Custom thresholds (defaults match Google's official Core Web Vitals thresholds: LCP 2500/4000 ms, INP 200/500 ms, CLS 0.1/0.25).
slackWebhookUrl	Send the CWV digest to a Slack incoming webhook (encrypted secret).
googleApiKey	Optional custom Google Cloud API key (X-goog-api-key). Recommended in audit mode for lists above ~200 pages.
maxPages, maxRunSecs	Cap on unique pages per run (default 200, max 2,000) and soft run time budget in seconds.

Monitoring and scheduling

Set syncKey and schedule the Actor. The Actor stores compact metric state (p75 values, ratings, assessment, category scores and timestamp — never full Lighthouse reports) in your account's key-value store cwv-monitor-sync-<key> (entries kept for 180 days, up to 50,000 page × device pairs).

• Schedule daily or weekly, not hourly. Google updates the 28-day CrUX dataset once a day around 04:00 UTC, and CrUX History once a week on Mondays. Running more than once a day returns the same CrUX numbers.
• Unchanged checks cost less. When a monitored page's CrUX metrics have not changed since the previous check (changeType: "unchanged"), it is billed at the discounted monitor-check-unchanged rate instead of field-check or audit. Set onlyChanges: true to omit unchanged rows from the dataset so downstream automations receive only new or changed rows.

Output

Each row in the default dataset is one page × device check:

{
    "url": "https://www.wikipedia.org/",
    "finalUrl": "https://www.wikipedia.org/",
    "origin": "https://www.wikipedia.org",
    "device": "phone",
    "mode": "audit",
    "status": "ok",
    "statusReason": "ok",
    "statusDetail": null,
    "fieldSource": "url",
    "cwvAssessment": "passed",
    "collectionPeriod": { "firstDate": "2026-10-06", "lastDate": "2026-11-02" },
    "lcpP75Ms": 1301,
    "inpP75Ms": 100,
    "clsP75": 0,
    "fcpP75Ms": 1170,
    "ttfbP75Ms": 546,
    "fieldMetrics": {
        "lcp": { "p75": 1301, "good": 0.9161, "needsImprovement": 0.0538, "poor": 0.0302, "rating": "good" },
        "inp": { "p75": 100, "good": 0.9346, "needsImprovement": 0.0519, "poor": 0.0136, "rating": "good" },
        "cls": { "p75": 0, "good": 0.986, "needsImprovement": 0.0115, "poor": 0.0025, "rating": "good" },
        "fcp": { "p75": 1170, "good": 0.8714, "needsImprovement": 0.0822, "poor": 0.0464, "rating": "good" },
        "ttfb": { "p75": 546, "good": 0.8471, "needsImprovement": 0.109, "poor": 0.0439, "rating": "good" },
        "rtt": { "p75": 134, "good": 0.451, "needsImprovement": 0.4693, "poor": 0.0796, "rating": null }
    },
    "urlMetrics": { "...": "same shape as fieldMetrics" },
    "originMetrics": { "...": "same shape as fieldMetrics" },
    "trend": {
        "weeksAvailable": 25,
        "firstDate": "2026-05-16",
        "lastDate": "2026-10-31",
        "direction": "stable",
        "lcp": { "firstP75": 1295, "lastP75": 1297, "changePercent": 0.2, "direction": "stable" },
        "inp": { "firstP75": 99, "lastP75": 100, "changePercent": 1, "direction": "stable" },
        "cls": { "firstP75": 0, "lastP75": 0, "changePercent": 0, "direction": "stable" },
        "series": [
            { "endDate": "2026-10-31", "lcpP75": 1297, "inpP75": 100, "clsP75": 0, "fcpP75": 1168, "ttfbP75": 544 }
        ]
    },
    "labPerformanceScore": 96,
    "labCategories": { "performance": 96, "accessibility": 99, "bestPractices": 81, "seo": 100 },
    "labMetrics": { "lcpMs": 1120, "cls": 0, "tbtMs": 15, "fcpMs": 890, "speedIndexMs": 1150, "ttfbMs": 120 },
    "recommendations": [
        {
            "id": "document-latency-insight",
            "title": "Document request latency",
            "displayValue": "Est savings of 850 ms",
            "savingsMs": 850,
            "savingsBytes": null
        }
    ],
    "lighthouseWarnings": [],
    "lighthouseVersion": "13.6.0",
    "changeType": "first-check",
    "changedMetrics": [],
    "previousCheckedAt": null,
    "alerts": [],
    "source": "Chrome UX Report, Google, CC BY 4.0; Google PageSpeed Insights (Lighthouse)",
    "checkedAt": "2026-11-04T06:30:00.000Z"
}

The dataset provides four table views: Overview, Field data (CrUX), Lab audit (Lighthouse) and Alerts & Monitoring.

Run report (OUTPUT)

Stored in the default key-value store as OUTPUT:

• cwvByDevice: counts of passed, failed and insufficientData per device (phone, desktop, tablet).
• domainsReport: per-origin summary with origin CWV assessment, p75 metrics, weekly trend direction, and how many checked pages on that domain passed or failed.
• worstPages: up to 10 slowest pages sorted by CWV failure and LCP p75.
• degraded, improved, failedPages: pages that worsened or improved since the previous run under syncKey.
• noFieldDataUrls: URLs that had neither URL-level nor origin-level CrUX data.
• statusCounts, statusReasons: breakdown of ok, no-field-data, http-404, blocked-lab, dns-error, timeout and lab-error.
• requests: request counts and rate-limit wait times for HTTP pre-checks, CrUX API and PageSpeed Insights API.

Measured

Measured against live Google Chrome UX Report API and PageSpeed Insights v5 API (2026-11-04, 147 CrUX queries and 120 PSI audits across 32 diverse domains and subpages):

What	Result
CrUX API response time	Median 0.45 s; 0 server errors across 147 queries
CrUX coverage (32 diverse URLs: major, mid, small)	72% had URL-level data; 94% had origin-level data (fieldSource: "origin" covers subpages)
CrUX History availability	100% of URLs and origins with daily CrUX data also had full 40-week history
CrUX API vs PageSpeed Insights loadingExperience	90 / 90 identical p75 values
PageSpeed Insights audit duration	Median 16.7 s, p90 61.7 s, max 110.5 s (heavy sites like github.com can exceed 100 s)
Lighthouse score spread across 3 consecutive runs	Median 4.5–5 pts, max 17 pts (mobile) and 28 pts (desktop) — why alerts use CrUX, not lab
Edge cases caught automatically	HTTP 404 pages (PSI scores them 90–100 with a warning), bot-blocked labs (nytimes.com), HTTP redirects

Pricing

Pay per event; platform compute usage is included.

Event	Price (FREE / BRONZE)	SILVER / GOLD	When
run-start	$0.0045	$0.0035 / $0.003	Once per run, charged only when the run delivers at least one paid check
field-check	$1.95 per 1,000	$1.50 / $0.99	One page × device check with CrUX field data in field mode
audit	$3.90 per 1,000	$3.20 / $2.50	One page × device Lighthouse + CrUX audit in audit mode
monitor-check-unchanged	$0.95 per 1,000	$0.75 / $0.50	One monitored page × device check under syncKey whose CrUX metrics did not change

Typical job	Estimated cost (BRONZE)
100 URLs × 2 devices in field mode (200 checks + 25-wk history)	~$0.39
50 URLs × 2 devices in audit mode (100 Lighthouse + CrUX)	~$0.39
Daily monitor of 100 URLs × 2 devices (all unchanged)	~$0.19
Pages without CrUX data (no-field-data), 404s, DNS or lab errors	free ($0)

You never pay for URLs without CrUX data in field mode (no-field-data), HTTP 404 pages, non-existent domains (dns-error), pages that block Lighthouse (blocked-lab), timeouts or a run that delivers no billable results (no start fee either).

Limitations

• Low-traffic pages have no URL-level CrUX data. Google includes a URL or origin in the Chrome UX Report only when it is publicly discoverable and has sufficient real Chrome traffic. Product subpages and small sites often fall back to fieldSource: "origin", and very small sites return statusReason: "no-field-data" (free of charge).
• CrUX updates once a day (~04:00 UTC) and reflects a 28-day rolling average (~2 days behind real time). CrUX History updates once a week on Mondays. Running field checks more than once a day returns the same numbers.
• Lighthouse lab scores vary between runs. In 3 back-to-back PSI runs on the same URL, Lighthouse performance scores varied by a median of 4.5–5 points and up to 28 points on complex pages. Treat labCategories and labMetrics as diagnostic snapshots, and rely on CrUX field metrics for regression alerts.
• Google API quotas. Google enforces 150 queries/min per Cloud project on the CrUX API (~3,000–4,500 pages/hour in field mode) and 30 queries/min (25,000/day) per project on the PageSpeed Insights API (~900 page × device audits/hour in audit mode). When multiple large runs share the built-in key, automatic pacing and 429 retries slow the run down rather than failing it (OUTPUT.requests reports any wait time). For audit runs over ~200 pages, supply your own free Google Cloud API key in googleApiKey for dedicated project quota.
• Some sites block Lighthouse or time out. Sites with strict bot protection (e.g. nytimes.com) return HTTP 403 to Lighthouse (statusReason: "blocked-lab"), and very heavy pages (e.g. github.com) can return HTTP 500 or time out after 150 s (statusReason: "lab-error"). In both cases, CrUX real-user metrics are still returned when available, and the failed lab audit is not charged.

Data license and attribution

Field performance data in fieldMetrics, urlMetrics, originMetrics and trend is sourced from the Chrome UX Report, Google, licensed under the Creative Commons Attribution 4.0 International License (CC BY 4.0). Every dataset row carries this attribution in the source field.

Not affiliated with Google

This is an independent tool by DataRelay, not affiliated with, endorsed by, or sponsored by Google LLC.

Usage statistics

The Actor sends anonymous run statistics to the developer (input options, counts of statuses and assessments, Google API request counts, duration). Customer URLs, domains and your Apify user id are never sent — only keyed HMAC-SHA256 hashes used to count repeated targets and distinct users. Custom API keys (googleApiKey) and Slack webhook URLs are never logged or transmitted.

Changelog

• 0.1 – Initial release: CrUX API + 40-week CrUX History, PageSpeed Insights v5 (Lighthouse), HTTP redirect & 404 pre-checks, sitemap discovery, cross-run monitoring (syncKey) and Slack alert digests.