Frequently asked questions

Because slow pages lose people, and you rarely find out. Nobody files a complaint about a three-second load — they leave, and the analytics show a bounce with no reason attached.

There is a large body of published research linking load time to abandonment and conversion, mostly from retailers measuring their own traffic. The honest summary is that the direction is well established and the magnitude depends entirely on your audience, your device mix and what you are asking them to do. Anyone quoting you a single percentage is selling something.

The argument this tool actually rests on is narrower: your score moves every time you deploy, and without a record you will not notice the day it dropped.

Less than most articles about it suggest. Google's page experience signals — which include the Core Web Vitals — are a real ranking input, but a weak one: they act as a tiebreaker between results of otherwise comparable relevance. Google has said repeatedly that there is no speed penalty, and that relevant content on a slow page still outranks thin content on a fast one.

So do not buy a monitoring tool to move up the rankings. Buy one because the people who already reach your site are waiting on it, and because a score that silently slid from 95 to 71 over four deploys is worth knowing about.

Google Lighthouse is an open-source, automated tool developed by Google to audit web pages. It runs directly in your browser (built into Chrome DevTools) or via command line, and generates a report on how well your page performs.

Lighthouse tests your page in a simulated environment and measures things like loading speed, accessibility, SEO basics, and adherence to web development best practices.

PageSpeed Insights (PSI) is Google's web-based tool that uses Lighthouse under the hood. When you enter a URL on pagespeed.web.dev, Google runs a Lighthouse audit on their servers and returns the results.

The key difference: PSI also includes real-world data from the Chrome User Experience Report (CrUX), showing how actual visitors experience your site. Lighthouse alone only provides lab data from simulated tests.

LighthouseRobot uses the official PageSpeed Insights API, which means you get the same Lighthouse scores you'd see on pagespeed.web.dev — but automated and tracked over time.

I chose the PSI API because it's reliable, free, and runs tests from Google's infrastructure rather than requiring my own testing servers. This keeps the service simple and the results consistent.

The Performance score measures how fast your page loads and becomes interactive. It's based on several metrics including:

• First Contentful Paint (FCP) — When the first content appears
• Largest Contentful Paint (LCP) — When the main content is visible
• Total Blocking Time (TBT) — How long the page is unresponsive
• Cumulative Layout Shift (CLS) — Visual stability during loading
• Speed Index — How quickly content is visually displayed

This is often the most volatile score as it depends on server response times, network conditions, and resource optimization.

The Accessibility score measures how usable your site is for people with disabilities. It checks for common issues like:

• Missing alt text on images
• Insufficient color contrast
• Missing form labels
• Incorrect heading hierarchy
• Missing ARIA attributes
• Keyboard navigation issues

A high accessibility score means your site is usable by more people and may also be legally required in many jurisdictions.

The Best Practices score checks whether your site follows modern web development standards and security practices:

• Using HTTPS
• Avoiding deprecated APIs
• No browser errors in the console
• Correct image aspect ratios
• Proper charset declaration
• Safe use of JavaScript

This score helps ensure your site is secure, reliable, and future-proof.

The SEO score checks basic search engine optimization factors that help your pages get discovered:

• Valid meta descriptions and titles
• Crawlable links
• Valid robots.txt
• Legible font sizes
• Proper use of hreflang tags
• Valid structured data

Note: This checks technical SEO basics, not content quality or backlinks.

LighthouseRobot uses Google's official color thresholds:

• Red (0–49) — Poor. Needs significant improvement.
• Yellow (50–89) — Needs improvement. Room for optimization.
• Green (90–100) — Good. Meets best practices.

You should — it is free, it takes ten seconds, and it is the same engine this runs on. For a one-off check on one page, opening pagespeed.web.dev is the right answer and nothing here beats it.

What it cannot do is happen without you. Four things follow from that:

• It runs whether or not you remember. Every site you add is audited every 24 hours, three runs, median stored. The scores that matter are the ones taken on the deploy you were not thinking about.
• One run is not a measurement. Lighthouse swings 10–15 points between identical runs. A single number tells you where the noise landed that minute; a line over 30 days tells you what your site does.
• It is showable, and it keeps being true. A screenshot proves nothing — anyone can make one. A share link and a badge point at a live record that is re-measured every day and can get worse. Send a client the link once and it stays current on its own; embed the badge and it updates without you.
• It works like a quality mark you did not buy. Trust seals are usually purchased and say what the buyer paid for. This one is a measurement: Google produces the number, it is public, it carries the date it was taken, and it turns amber or red when the site does. You cannot set it to “good” — which is the only reason showing it means anything.
• Your AI assistant can read it. The MCP server puts the whole history behind a question: ask Claude or Cursor what regressed since last week and it answers from your own measurements, without you opening a browser tab or pasting a URL anywhere.

Put plainly: PageSpeed Insights answers “how is this page right now?” This answers “what has been happening to my sites, and can I show someone?”

LighthouseRobot uses Google's official PageSpeed Insights API to audit your websites. When an audit runs, I make three separate API calls and take the median score for each metric.

This “median-run strategy” significantly reduces score variability. A single Lighthouse test can vary by 10–15 points for the same page; taking the median reduces this to approximately 3 points.

All monitored sites are audited automatically every 24 hours. Each audit runs 3 tests and takes the median score for stability.

The Total Score displayed on LighthouseRobot is a simple average of all four Lighthouse scores (Performance + Accessibility + Best Practices + SEO) divided by 4.

This gives you a quick at-a-glance indicator of your overall website health, while the individual scores show you exactly where to focus improvements.

Lighthouse scores — especially Performance — can vary between runs due to:

• Server response time variations
• Network conditions
• Third-party scripts loading differently
• CDN cache states
• A/B tests or dynamic content

LighthouseRobot uses daily averages and median-run scoring to smooth out these fluctuations and give you meaningful trend data.

Every site is measured with desktop settings, and that is what the leaderboard, the badges and the share pages show. Desktop gives the more stable and consistent scores.

A mobile measurement now runs alongside it. It is kept as its own history, so the two never overwrite each other, and the board can be switched between them. It is still limited to a small number of accounts while it is being tested, so it is not switched on for everyone yet.

A Leaderboard is a collection of websites you want to compare. Your sites are automatically ranked by their Total Score, creating a competitive view of your portfolio's performance.

Each account can have one leaderboard with up to 10 monitored sites.

The Monitor detail page shows comprehensive data for a single website:

• Current scores (daily average)
• Historical score chart showing trends over time
• Core Web Vitals plus the two supporting lab metrics, with visual range indicators
• Resource overview (total requests, page size)
• Individual audit history
• Quick actions like running a new audit or sharing

This view helps you track how your optimization efforts impact scores over time.

Core Web Vitals are the three metrics Google treats as thresholds for real-world user experience: LCP (loading), INP (interactivity, which replaced FID in March 2024) and CLS (visual stability). They are field metrics — measured from real visits to your site, not from a test.

LighthouseRobot runs a lab test and, where Google has enough data for your site, shows the field measurement beside it. That distinction matters, so here is exactly what the lab test gives you and what it cannot:

• Largest Contentful Paint (LCP) — a Core Web Vital. How fast the main content loads. Good: under 2.5s
• Cumulative Layout Shift (CLS) — a Core Web Vital. Visual stability while the page loads. Good: under 0.1
• Total Blocking Time (TBT) — not a Core Web Vital. It is the lab stand-in for INP: how long the main thread was blocked, which is what makes a page feel unresponsive. Good: under 200ms
• First Contentful Paint (FCP) — not a Core Web Vital. When the first pixel of content appears. Good: under 1.8s

No lab test can measure INP — it needs real user interactions over time, and a scheduled audit produces none. TBT is the closest a lab test gets, and it correlates well. So since August 2026 every monitor page and every shared report carries a second block, Field data, which reads LCP, INP and CLS straight from Google's Chrome User Experience Report — the same source Google uses for ranking. Your real INP is there when it exists.

It does not always exist. Google only publishes that report for sites with enough Chrome visitors, so smaller and newer sites see a note instead of numbers. That is a statement about traffic, not about quality: the lab measurement above it runs either way.

On rankings: page experience is one weak signal among many, and Google does not penalise a site for being slow. A fast site is worth building for the people using it; treat any ranking effect as a bonus rather than the reason.

You can generate public share links for both your leaderboards and individual monitors. Anyone with the link can view the shared content without needing an account.

This is perfect for sharing performance reports with clients, posting comparisons on social media, embedding in presentations, or showcasing your optimization work.

Yes. When you enable sharing for a monitor, you get an embeddable SVG badge that shows all four Lighthouse scores as visual rings. The badge updates automatically with your latest scores.

Simply copy the HTML code from the Share modal and paste it into your website. The badge uses the official Lighthouse colors (green/orange/red) and links back to your public share page.

This is great for showing visitors that you care about performance, or for adding a trust signal to your portfolio.

Yes, by default all your data is private. Only you can see your dashboard, monitors, and audit history when logged in.

Share links are opt-in — you decide what to make public. You can deactivate share links at any time.

After each audit, LighthouseRobot compares the new total score against the previous one for that monitor. If it has fallen by at least your chosen threshold, you get one email.

• You pick the threshold: 5, 10 or 15 points. Below it, nothing is sent — Lighthouse scores naturally wobble by a few points between runs, and an alert on that noise is worse than no alert at all.
• One email per monitor per day, at most. A site having a bad week cannot fill your inbox.
• The comparison is against the previous audit, not against a long-term average, so a slow slide across many days may never cross the threshold on any single day. The 30-day chart is what catches that.
• A monitor's first audit never alerts — there is nothing to compare it against.

Plain text, deliberately: the site name and URL, the old and new total score, the size of the drop, and a direct link to that monitor. No tracking pixel, no images, no HTML that renders differently in every client. The homepage shows the actual message.

No. Alerts are off until you switch them on under Account Settings, and while they are off your email address is never sent to our email provider at all. Once enabled, your address, the monitor name and its URL are transmitted to Resend so the message can be delivered — see the Privacy Policy for the detail.

MCP (Model Context Protocol) is an open standard that allows AI assistants like Claude to connect to external data sources. LighthouseRobot provides an MCP server that lets you query your performance data directly from your AI assistant.

Instead of logging into the dashboard, you can simply ask: “How is my website performing?” or “What should I optimize on example.com?” and get instant answers.

LighthouseRobot's MCP server works with any MCP-compatible client, including:

• Claude Desktop — Anthropic's desktop app for macOS and Windows
• Cursor — The AI-powered code editor
• Other MCP clients — Any tool that supports the MCP protocol

Once connected, you can ask questions like:

• “Show me my dashboard overview”
• “List all my monitored websites”
• “How is example.com performing?”
• “What should I optimize on my website?”
• “Generate a performance report for the last 7 days”
• “Which websites improved or declined this week?”
• “Show me the ranking history for my site”

The AI can provide actionable improvement suggestions based on your PageSpeed data.

The MCP server provides these tools:

• list_dashboards — Every dashboard on your account, with its leaderboards and monitor counts
• get_dashboard — Overview with average scores, across every dashboard or one you name
• list_monitors — All monitored websites with current scores
• get_monitor — Detailed data including Core Web Vitals
• get_improvement_suggestions — Actionable optimization tips from PageSpeed
• get_performance_report — Trend analysis over 7/14/30/90 days
• get_ranking_history — How your ranking changed over time
• get_winners_losers — Best and worst performers in your portfolio

To connect your AI assistant, you need an API token:

1. Log into your LighthouseRobot account
2. Go to Account Settings → API Tokens
3. Click “Generate New Token”
4. Give it a name (e.g., “Claude Desktop”)
5. Copy the token immediately — it won't be shown again!

Keep your token secure. Anyone with the token can access your LighthouseRobot data.

Add the following to your Claude Desktop configuration file:

1{2  "mcpServers": {3    "lighthouserobot": {4      "command": "npx",5      "args": [6        "-y",7        "mcp-remote",8        "https://lighthouserobot.com/api/mcp",9        "--header",10        "Authorization: Bearer YOUR_API_TOKEN"11      ]12    }13  }14}

Config file location:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

You do not have to find the file by hand: in Claude Desktop, open Settings → Developer → Edit Config and it opens the right file, creating it if it does not exist yet.

Then quit Claude Desktop completely and reopen it — a window reload is not enough, the app reads this file at startup. You should see LighthouseRobot in your available tools.

Create a file called .cursor/mcp.json in your project folder:

1{2  "mcpServers": {3    "lighthouserobot": {4      "url": "https://lighthouserobot.com/api/mcp",5      "headers": {6        "Authorization": "Bearer YOUR_API_TOKEN"7      }8    }9  }10}

Restart Cursor after adding the configuration.

Yes. All communication happens over HTTPS, and your API token is hashed (SHA-256) before being stored in the database — we never store the plain token.

You can revoke tokens at any time from your Account Settings, and create separate tokens for different devices or applications.

Just me! LighthouseRobot is a personal side project. I built it because I wanted a simple way to track Lighthouse scores for multiple websites without running manual tests all the time.

There's no company, no team, no investors — just one person who enjoys building useful tools.

Yes, completely free. No credit card required, no trial period, no hidden costs. I built this for myself and decided to share it. Enjoy!

Each account includes:

• 1 leaderboard
• Up to 10 monitored websites
• Automated audits every 24 hours, 3 runs each
• Unlimited share links
• 30-day score history