PageSpeed Insights — Google Lighthouse Audits

Batch PageSpeed audits without an API key

For developers, SEO teams, and site owners checking pages after a release, this Actor returns one Google Lighthouse report per URL and device. Supply public URLs and export performance, accessibility, best-practice, and SEO scores with lab measurements and prioritized fixes.

The default engine runs Google Lighthouse in Chrome on Apify. No Google account, API key, provider subscription, or browser setup is required. Audit conditions differ from Google's hosted PageSpeed Insights service, so scores can differ. Each result identifies its source.

What you get

Field	Meaning
url, finalUrl, strategy	Requested page, resolved page, and mobile/desktop device
performanceScore, accessibilityScore, bestPracticesScore, seoScore	Scores from 0 to 100; unselected categories can be null
lcpMs, fcpMs, tbtMs, speedIndexMs, ttfbMs	Available Lighthouse lab measurements in milliseconds
cls	Unitless lab layout-shift score
opportunities	Top optimization recommendations ranked by estimated time savings, then byte savings
failedAudits, failedAuditCount	Failed Lighthouse checks and their total count before detail limits
fieldData.url, fieldData.origin	Separately labeled field measurements when supplied by optional Google API mode
source, lighthouseVersion, fetchTime, warnings	Audit provenance, engine version, measurement time, and warnings

The default keyless engine provides lab data. Its CrUX field measurements are marked available: false. Lab TBT is distinct from field INP; absent field data does not establish a Core Web Vitals pass.

Start with your important pages

{
  "urls": ["https://example.com/"],
  "strategies": ["mobile", "desktop"],
  "categories": ["performance", "accessibility", "best-practices", "seo"],
  "maxAudits": 2
}

Run the same page on both devices to identify mobile-specific issues. Compare results after a deployment, then assign the largest supported optimization opportunities to your engineering backlog. For broader technical crawl checks after performance triage, use a separate site crawler; this focused workflow audits exactly the URLs you submit.

Example result

This shortened example illustrates the persisted report shape; other required fields remain present in the actual Dataset.

{
  "url": "https://example.com/",
  "finalUrl": "https://example.com/",
  "strategy": "mobile",
  "performanceScore": 99,
  "accessibilityScore": 96,
  "bestPracticesScore": 100,
  "seoScore": 80,
  "lcpMs": 795.43,
  "cls": 0,
  "source": "Google Lighthouse (Apify-hosted Chrome)"
}

Input controls and supported pages

• urls: 1–100 public HTTP(S) URLs. Exact duplicates and fragments are removed. Embedded credentials, private networks, local hostnames, and custom ports are rejected.
• strategies: mobile by default; choose desktop or both. Each URL/device pair is a separate audit.
• categories: performance, accessibility, best-practices, and SEO by default. Selecting fewer categories does not multiply the audit fee.
• maxAudits: defaults to 20, maximum 200. Failed attempts also consume this work limit. Input URL order and selected device order determine processing order.
• maxOpportunities: 0–30, default 10. maxFailedAudits: 0–100, default 30. Set either to zero for compact reports.
• locale: audit text language, default en.
• requestTimeoutSecs: per-audit deadline, 10–120 seconds, default 90. maxDurationSecs: run work deadline, 30–1800 seconds, default 900. Leave additional Apify run time for finalization.

Each keyless audit uses a fresh browser session, one bounded attempt, and a 20 MiB browser-network limit. Login-required pages, paywalls, security challenges, and very large pages may be unavailable. Use stable, publicly accessible pages and allow enough run time for the URL/device combinations you request. The Actor audits supplied URLs rather than crawling sitemaps or discovering links.

Optional Google-hosted API mode

Advanced callers may enter their own enabled Google PageSpeed key in the secret apiKey field to explicitly select Google's hosted API. This optional key is excluded from Dataset results, output, and logs. Keyless Lighthouse remains the default whenever the field is omitted.

Google API mode can include URL/origin CrUX field data when Google returns it; Google has announced its removal from this API. Quota/configuration errors stop additional API work. maxRetries controls only transient errors in optional API mode, with one extra attempt by default and at most two. The default keyless engine requires no API key and has no external API request fee.

Pricing and cost control

The Actor uses Pay per event: $0.003 per successful, validated, saved URL/device report ($3 per 1,000 reports), plus caller-paid Apify platform usage for compute, storage, and transfer. Check the live Pricing tab for the current effective fee.

Both devices on one URL produce two result events. All selected Lighthouse categories are included in each event. There is no startup fee. Failed audits, invalid inputs, Google quota/configuration errors, and retries have no result-event fee; platform usage may still accrue. Optional Google API quota belongs to the caller's Google project.

Use Apify's maximum run spending setting together with maxAudits. The Actor checks result affordability before starting another audit, preserves the last affordable report, and reports unfinished work explicitly. The event limit covers Actor result charges; platform usage is an additional cost. Browser audits generally use more compute than a simple HTTP lookup.

Results, partial runs, and terminal records

Successful reports go to the default Dataset, where Apify offers JSON, CSV, and Excel export. OUTPUT summarizes outcome and saved rows; RUN_SUMMARY includes attempts, failures, skipped work, source, and acknowledged named-event counts.

COMPLETE means the scheduled work finished. PARTIAL preserves useful results when a page fails or a work/spending limit stops the run. INVALID_INPUT explains a user-correctable request. If valid advertised work produces no useful report because all routes fail, the run ends FAILED with UPSTREAM_FAILED. Missing engine/deployment configuration is CONFIG_ERROR and also fails honestly.

The serialized Dataset/billing writer validates each report before persistence and billing. Interrupted or uncertain effects require reconciliation before automatic continuation; storage and billing are separate effects, and SDK acknowledgement is distinct from invoice settlement.

Call from the Apify API

curl --request POST \
  'https://api.apify.com/v2/acts/khadinakbar~google-pagespeed-insights/runs' \
  --header "Authorization: Bearer $APIFY_TOKEN" \
  --header 'Content-Type: application/json' \
  --data '{"urls":["https://example.com/"],"strategies":["mobile","desktop"],"maxAudits":2}'

For AI agents and MCP clients: “Audit these public URLs with khadinakbar/google-pagespeed-insights on mobile and desktop, without an API key. Wait for the run, read OUTPUT and RUN_SUMMARY, retrieve every Dataset report, and summarize the largest supported fixes with URL, device, source, measurement time, and actual run cost.” The Actor's core acceptance is its batch API; client-specific hosted MCP behavior must be checked in your chosen client.

Builder's note and responsible use

Fresh isolated Chrome sessions keep URL/device audits independent. Redirects and subresources pass through a public-address-only proxy that validates and pins DNS destinations; this also prevents an audited page from reaching private services through its resources. Scores remain point-in-time measurements that can vary with network and page changes.

Audit public pages you own or are authorized to assess. Respect applicable site terms and access controls. Results are technical measurements, not ranking guarantees or a guarantee of legal compliance. Google Lighthouse is Apache-2.0 software. This independent Actor is not endorsed by Google.