Changelog
What's new
Every update to LighthouseRobot since the first dashboard — newest first, nothing rewritten after the fact.
Registered with Google Search Console
- Improvementlighthouserobot.com is now verified with Google Search Console, so how the public pages — leaderboards, reports, badges — show up in Google Search can be measured instead of guessed. The proof is one meta tag on every page; nothing changes for visitors
Straight to the live pages
- ImprovementEach chapter on the home page now links first to a real, public page — this site’s own audit log, a live leaderboard, a live report — and only second to the explanation on the product page. Before, the first link led to the product page, where the link to the real thing waited behind a second click
- ImprovementThe “New · …” link under the hero is gone. It was typed by hand and had said “Rebuilt the public site” through five later releases, and it competed with the two links that matter there: start free, or open a live leaderboard. What is new lives on this page, one click away in the navigation
- ImprovementThe line under the opening statement that pointed to the FAQ is gone. The FAQ is one click away in the navigation and the footer, and the statement reads better without a footnote
- ImprovementThe audit log on public monitor pages is folded by default. Thirty open rows pushed everything below them off the screen. The line that matters stays visible — how many audits, since when — and the rows open on a click into a box that scrolls on its own, with the column headers pinned. A link that points straight at the log opens it
Alerts that actually arrive, and a stricter door for assistants
- FeaturePublic monitor pages now carry an audit log: the last 30 measurements, newest first, with the time in UTC and all four scores plus the total — and how many audits stand behind the page since the monitor was created. The rings and the chart are summaries; this is the ledger behind them, for anyone who wants to check the claim row by row. A category PageSpeed could not measure shows as a dash, never as a zero
- FeatureThe badge has a JSON twin. Swap .svg for .json on any badge URL — /api/badge/<token>.json, with ?strategy=mobile as before — and you get the same numbers as data: the four scores and the total, a good / average / poor rating per category on Lighthouse’s own thresholds, the date they were measured, how long the site has been monitored, and links to the page and the SVG. Readable from other sites in the browser, so you can draw your own badge in Framer, Webflow or an Astro component instead of embedding ours
- FixScore-drop alerts were not being delivered. The sender gave up quietly when its mail configuration was incomplete, and the record of “alert sent” was written regardless — so the accounts that had turned alerts on got nothing, and nothing said so. That is fixed: a send that fails is now an error, and an alert is only recorded once the mail provider has confirmed it, with the provider’s receipt stored alongside. The daily pipeline heartbeat to the operator had the same gap and the same fix
- FixThe MCP server has tightened how it checks credentials. Tokens from the shelved ChatGPT integration are no longer accepted, and the leftover pieces of that integration are closed. API tokens generated in Account Settings — the way Claude and Cursor have always connected — are unaffected
A reference line on the home page
- FeatureThe home page now shows what every monitored site currently scores, averaged — Performance, Accessibility, Best Practices and SEO, with a Desktop/Mobile switch. A score on its own says little: 82 is respectable for performance and would be poor for SEO, and nothing on the page told you which. The figures come from the last three audits of each site and refresh hourly
- ImprovementMonitor pages now say where each number comes from. “Core Web Vitals” is “Laboratory data — Google’s synthetic Lighthouse test”, “Real user data” is “Field data — Google’s Chrome User Experience Report”, and page weight now stands on its own below both rather than under a heading that named Lighthouse as its source. Lab and field figures for one page routinely disagree — one mobile run recorded a lab LCP of 13.5s where real visitors saw 1.2s — and until now nothing on the page said which was which
- FixSharing any page of this site showed the wrong picture. The preview card was a screenshot of the home page as it looked on New Year’s Eve — an “Early Beta” chip, a headline that has not existed since the August rebuild — and it was the wrong shape, 1755×1671 where 1200×630 was declared. The seven subpages carried no preview image at all. They now all render the live card: this site’s own current scores, from the same daily audit every monitored site gets. Sharing a link is the one moment the site travels without us, and it had been advertising a page that is gone
- ImprovementThe FAQ answers the question it never answered: why not just run PageSpeed Insights yourself? Because it only happens when you remember, one run swings 10–15 points so a single number is noise, and a screenshot proves nothing — a share link and a badge point at a live record that is re-measured daily and can get worse. It works like a quality mark you did not buy: Google produces the number, it carries the date it was taken, and you cannot set it to good
- FeatureShared leaderboards can be switched between Desktop and Mobile, the same way your own board can. Desktop stays the default and the ranking axis — mobile scores sit systematically lower, so making them the basis would resort every board without anything having changed on those sites. The choice travels in the URL, so a shared mobile view is itself linkable. Boards with no mobile measurements show no switch at all: a control whose only outcome is a table of dashes is worse than none
- ImprovementAll 34 answers here now carry FAQPage structured data, and the home page has its own title and description instead of inheriting a line that still described the product as it was pitched in December. There is also an llms.txt now, stating plainly what this service measures and the five things it deliberately does not claim
- ImprovementEach feature on the home page links to its own chapter, and the FAQ is reachable from the paragraph that raises the question instead of only from the nav
- FixThe MCP server returned nothing at all to any account with more than one dashboard. Every one of its tools resolved your dashboard with a query that treats a second one as an error rather than a list, and the error was never read — so an account with four dashboards and 32 monitors got an empty answer from all seven tools, with no failure shown. It now covers every dashboard on the account by default, and a new list_dashboards tool names them; the tools that report on a set of monitors accept a dashboard_id to narrow it
- FixA revoked or expired API token was answered with a server error instead of an authentication error, so an assistant retried instead of asking you to sign in again. Both now return 401
The badge now says when it was measured
- ImprovementThe embeddable badge carries the date of its last successful measurement next to Desktop or Mobile — "Desktop · Checked Aug 29, 2026". Nothing greys out or turns red when audits stop: a date that reads eight months ago is plain enough on its own, and an outage on our side should never make your site look broken on your own page
- ImprovementEvery line on this page now says whether it is a fix, an improvement or a new feature
- ImprovementThe home page says what the daily score is for. The headline is unchanged — every site you ship, scored every day — but the line under it now names the point of that: the numbers you show are measured, not claimed. It is the same product; what it is good for is stated rather than left to be worked out
- FixPublic share pages — both a single monitor and a leaderboard — were still topped by the old header bar: a logo and nothing else, no navigation, no way on to the rest of the site. They now carry the same header and footer as every other page. These are the only pages someone can reach without ever seeing lighthouserobot.com, so they were the worst place to look like somewhere else
- FeatureNew on every monitor page and on shared reports: Real user data. Alongside the lab measurement, monitors now show what actual visitors experienced over the last 28 days, taken from Google’s Chrome User Experience Report. The two genuinely differ — one mobile run recorded a lab LCP of 13.5s for a page whose real visitors saw 1.2s. It also brings INP, which a lab test cannot produce at all, because it needs real interactions to observe. Google only publishes this for sites with enough Chrome traffic, so smaller sites will see a note instead of numbers — that is about visitor counts, not quality
Mobile measurement, and the Agentic column that hid it
- FeatureThe board can be switched between Desktop and Mobile. Both runs are stored as separate histories, so one never overwrites the other, and desktop stays what the leaderboard, the badges and the share pages show. Mobile is still limited to a small number of accounts while it is being tested
- FixThe switch has been running since August 26 without being mentioned here, and the FAQ still said mobile testing "might be added in the future". Both are corrected
- FixOn the mobile board, the Agentic Browsing column showed "–" for every site — the same dash that means "not measured yet" — even where a mobile measurement existed. It now reads the value from the mobile run itself. That matters because the two runs genuinely differ: on 3 of the 29 sites measured both ways, the mobile pass finds one more readiness check applicable than the desktop pass does
Delete Account now deletes, and a new measurement for AI agents
- FeatureNew column in the leaderboard and a block on every monitor page: Agentic Browsing, the category Lighthouse added for how well a site can be read by an AI agent. It is shown as a fraction — 2/2, 0/3 — because that is what it is: a count of the readiness checks that pass, out of the ones that apply to your site. Different sites have a different number of applicable checks, so the totals are not directly comparable, and the ranking is unchanged
- FixA monitor whose audits keep failing is now marked as such in the leaderboard. The warning was built last week but never reached the table
- FixDeleting your account removes everything: dashboards, monitors, audit history, share links, API tokens and the login itself. It previously signed you out and kept the data, which is not what the button, the confirmation dialog or the privacy policy said
- ImprovementDashboard, account and login pages can no longer be embedded in a frame on another site. Public share pages and badges are unaffected, since being embedded is their purpose
- FixThe improvement suggestions in the MCP server work again — a storage change the day before had left them returning nothing
- ImprovementOne measurement path instead of two: manual audits now run the same parser, in parallel, as the scheduled ones, and finish in a third of the time
Rebuilt the public site, and corrected the score maths
- FixFixed a rounding error that had been quietly costing scores a point: the total was truncated instead of rounded, so 9,702 of 21,989 audits read one point below the truth. Every affected score in the history has been recalculated, and 8 of the 24 leaderboards with more than one site changed order as a result
- FixA category PageSpeed cannot measure now shows as "–" rather than counting as a zero. Two monitors had recorded a 20-point crash that never happened, on and off since July
- FixScore-drop alerts no longer fire when the two audits being compared did not measure the same categories
- ImprovementEverything before the login rebuilt: new type scale, a proper surface and line system, and one motion language across every page
- ImprovementFeature sections alternate side to side, with the product panels drawn as real interfaces rather than screenshots
- ImprovementScore-drop alerts now have their own section, showing the email the way it actually arrives
- FixCorrected what the site says about Core Web Vitals: LCP and CLS are the ones measured here, TBT stands in for INP, and FCP is neither — the FAQ now says so plainly
- FixRemoved several claims that were not true: this site does not score a perfect 100 (it is 99/100/96/100, and the badge in the footer proves it), failed audit runs are not retried, and the footer no longer reports a service status nothing was checking
- FixEvery page has its own social preview card — they all shared the homepage before
- FixPrivacy policy now discloses Resend, which receives your email address when score-drop alerts are switched on, and the Terms list matches it
- ImprovementAccessibility pass: 44px tap targets, a working focus ring and focus trap, and the site renders fully with JavaScript disabled or reduced motion on
- ImprovementLegible print styles for the imprint, privacy and terms pages
Score Drop Email Notifications
- FeatureGet notified by email when a monitor's total score drops significantly
- FeatureOpt-in toggle in Account Settings — no alerts unless you want them
- FeatureConfigurable threshold: choose 5, 10, or 15 points as your trigger
- Feature24-hour cooldown per monitor prevents alert fatigue from natural score fluctuations
- FeatureEmails include monitor name, old/new score, and a direct link to the details
MCP Server & API Access
- FeatureNew MCP (Model Context Protocol) server for AI assistant integration
- FeatureConnect Claude Desktop, Cursor, or other MCP-compatible clients to your LighthouseRobot data
- FeatureGenerate API tokens in Account Settings → API Tokens
- FeatureAvailable tools: get_dashboard, list_monitors, get_monitor, get_improvement_suggestions, get_performance_report, get_ranking_history, get_winners_losers
- FeatureAsk your AI assistant questions like "How is my website performing?" or "What should I optimize on example.com?"
- FeatureGet actionable improvement suggestions directly from your PageSpeed data
- FeatureGenerate performance reports with trend analysis and Core Web Vitals tracking
Claude Desktop Configuration:
{ "mcpServers": { "lighthouserobot": { "command": "npx", "args": [ "-y", "mcp-remote", "https://lighthouserobot.com/api/mcp", "--header", "Authorization: Bearer YOUR_API_TOKEN" ] } }}Config location:
~/Library/Application Support/Claude/claude_desktop_config.json(macOS)Social Media Previews
- FeatureDynamic Open Graph images for all share links
- FeatureShare links now show score previews on LinkedIn, Twitter, Slack, and other platforms
- FeatureMonitor shares display the total score ring plus all 4 category scores
- FeatureLeaderboard shares show top 5 websites with their rankings
- FeatureClear "Lighthouse Score" and "Performance Leaderboard" labels
- FeatureCall-to-action footer: "Track your website performance → lighthouserobot.com"
Example: How your share link looks on LinkedIn, Twitter & Co:
Embeddable Performance Badge
- FeatureNew embeddable SVG badge showing all 4 Lighthouse scores (Performance, Accessibility, Best Practices, SEO)
- FeatureBadge updates automatically with your latest scores
- FeatureCopy-paste HTML code directly from the Share modal
- FeatureUses official Lighthouse colors (green/orange/red based on score)
Live example from lighthouserobot.com:
Mobile Experience & Landing Page
- ImprovementImproved mobile leaderboard: Rank, Site, and Total Score visible without scrolling
- ImprovementAbbreviated column headers on mobile (Perf, A11y, BP) for better fit
- ImprovementTooltips show full site names when truncated
- FixHeader layout stacks properly on mobile devices
- ImprovementAdded live example links on landing page (Leaderboard & Apple.com detail view)
Landing Page Redesign & Monitor Sharing
- ImprovementRedesigned landing page with product screenshots
- ImprovementInteractive demo link to explore the dashboard
- ImprovementVisual split-screen sections highlighting key features
- FeatureShare single monitors with anyone via public link
- ImprovementVisual indicator shows when sharing is active (pulsing green dot)
- FeatureNew Core Web Vitals section on monitor detail page
- ImprovementVisual range bars show where your metrics fall (good/needs improvement/poor)
- ImprovementInfo tooltips explain what each metric measures
- ImprovementLarger, more prominent Total Score display
- FeatureResource overview showing total requests and page size
Mobile Improvements
- ImprovementBetter leaderboard display on smartphones
- ImprovementClickable website links in the dashboard are now more visible
- ImprovementMore consistent leaderboard design across views
Major Feature Update
- ImprovementNew landing page design with a fresh look
- FeatureAdded About page with personal story
- FeatureComprehensive FAQ page explaining Lighthouse & PageSpeed
- FeaturePublished Imprint and Privacy Policy
- FeatureShare links: Share your leaderboards with others
- FeatureRanking movements: See at a glance which websites improved or declined
- ImprovementMobile hamburger menu for better navigation
- ImprovementSEO improvements for better discoverability
- ImprovementAutomatic retry mechanism for connection issues
Automation & Live Data
- ImprovementDashboard now shows daily averages for more stable results
- FeatureAutomatic performance measurements every 8 hours (3x daily)
- ImprovementLive data updates without page reload
- ImprovementLast measurement timestamp visible
- ImprovementImproved navigation between monitors
Project Launch
- FeatureFirst version of LighthouseRobot
- FeatureDashboard with website overview
- FeatureColored score rings for quick overview
Start your own record.
Ten releases of history on this page. Yours starts with the first audit — every 24 hours, kept for 30 days.