Bulk Lighthouse & Core Web Vitals Audit with Budgets

What does the Bulk Lighthouse & Core Web Vitals Audit do?

It runs real Google Lighthouse, the same engine behind PageSpeed Insights and Chrome DevTools, on a list of pages or a whole sitemap, on mobile and desktop. Each page gets its performance, accessibility, best-practices, and SEO scores, its Core Web Vitals (LCP, CLS, TBT), and a pass/fail result against your performance budgets.

No Google API key and no PageSpeed quota. Lighthouse runs inside the Actor on its own Chrome. Because it runs on Apify, you can schedule it (for example, a nightly check after each deploy), call it from the API, or send failures to Slack, email, or Google Sheets.

Why use it?

• Audit a whole site at once instead of pasting URLs into PageSpeed Insights one by one.
• Catch regressions. Set budgets (for example, LCP ≤ 2.5 s, CLS ≤ 0.1, performance ≥ 90) and get a clear pass/fail list after every release.
• Prioritize fixes. Each failing page lists its biggest Lighthouse opportunities, such as "Reduce unused JavaScript (~2,730 ms)".
• Report to clients. The flat, one-row-per-page output drops straight into a spreadsheet or dashboard.
• Pay for problems only. With Save only pages that fail a budget, healthy pages are counted but not charged.

How to use it

1. Open the Input tab and paste your page URLs, or add your sitemap URL.
2. Choose Mobile, Desktop, or both.
3. Optional: adjust the budgets, or turn on Save only pages that fail a budget.
4. Click Start. Each audit takes about 15-40 seconds.
5. Open the Output tab. Use Scores and Core Web Vitals for the overview, or What to fix for the biggest wins.

Input

Field	What it does
URLs to audit	One page per line.
Sitemaps to expand	Audits every page in a sitemap or sitemap index (up to Max URLs).
Device	Lighthouse's official mobile emulation (throttled 4G on a mid-range phone), desktop, or both.
Lighthouse categories	Performance, accessibility, best practices, SEO. Fewer categories run faster.
Budgets	Minimum performance score, and maximum LCP, CLS, and TBT. Defaults use Google's "good" thresholds.
Save only pages that fail a budget	Skip saving (and paying for) healthy pages.
Max URLs / Timeout	A safety limit per run, and per audit.

{
  "sitemapUrls": ["https://example.com/sitemap.xml"],
  "device": "both",
  "lcpMaxMs": 2500,
  "clsMax": 0.1,
  "onlyFailures": true
}

Output

One row per page and device. You can download the dataset in various formats such as JSON, HTML, CSV, or Excel.

{
  "url": "https://example.com/pricing",
  "device": "mobile",
  "budgetPass": false,
  "failedBudgets": ["performance: 30 (budget ≥ 90)", "lcpMs: 19220 (budget ≤ 2500)", "tbtMs: 4967 (budget ≤ 200)"],
  "performance": 30,
  "accessibility": 94,
  "bestPractices": 96,
  "seo": 92,
  "lcpMs": 19220,
  "fcpMs": 3100,
  "cls": 0.02,
  "tbtMs": 4967,
  "speedIndexMs": 9800,
  "ttfbMs": 120,
  "totalBytes": 2854301,
  "requests": 112,
  "topOpportunities": ["Reduce unused JavaScript (~2730 ms)"],
  "runtimeError": null,
  "lighthouseVersion": "13.5.0",
  "checkedAt": "2026-09-24T17:00:00.000Z"
}

A SUMMARY record in the run's key-value store counts audits, passes, failures, errors, and saved rows.

Data fields

Field	Description
performance, accessibility, bestPractices, seo	Lighthouse category scores (0-100)
lcpMs, cls, tbtMs	Core Web Vitals lab metrics. TBT is Lighthouse's stand-in for INP.
fcpMs, speedIndexMs, ttfbMs	First Contentful Paint, Speed Index, and server response time
budgetPass, failedBudgets	Pass/fail against your budgets, with each failure spelled out
topOpportunities	Lighthouse's biggest estimated savings for the page
totalBytes, requests	Page weight and request count
runtimeError	Why an audit couldn't finish (for example, the page didn't load)

How much does it cost to run Lighthouse on many pages?

You pay per result: one result is one page on one device. With Save only pages that fail a budget, passing pages cost nothing. Set a maximum charge per run in the run options, and the Actor stops cleanly at that limit, keeping everything audited so far.

Tips

• Audits run one at a time on purpose. Parallel Lighthouse runs on one machine skew each other's performance scores, so results here are consistent from run to run.
• Lab data, not field data. Lighthouse measures a controlled, simulated load. Real-user field data (CrUX) can differ, and trends between runs are what matter most.
• Schedule it after deploys, and send failures to Slack with an Apify integration.

Use it from AI agents (MCP)

AI agents can find and run this Actor through the Apify MCP server.

• Claude, ChatGPT, or any MCP client: add https://mcp.apify.com?tools=creativefour/lighthouse-audit as a custom connector, and sign in to Apify when prompted.
• Claude Code, Cursor, VS Code, or Codex: run apify mcp install claude-code (swap in your client's name), then ask your agent to "run a Lighthouse audit of example.com with creativefour/lighthouse-audit".

FAQ and support

Is this the same as PageSpeed Insights? It's the same Lighthouse engine and the same scoring. PageSpeed Insights also shows Google's real-user field data, which this Actor doesn't include.

Why does a score change between runs? Network and server timing vary a little on every load. For decisions, compare a few runs or look at trends.

Can it audit pages behind a login? No. It audits public pages only.

Found a bug or need a feature? Open an issue on the Issues tab. Custom versions are available on request.