What the Lighthouse+ AI Bridge audit checks

Three checks, one verdict. Each tile is a primitive your AI agent can read alongside the full JSON payload.

Performance score

Runs Google PageSpeed Insights on a simulated mobile device and reports the 0-100 performance score: 90 and above is good, 50 to 89 needs improvement, below 50 is poor and sets the checker to warning.

Accessibility

Reports the Lighthouse accessibility score from the same run: color contrast, form labels, image alt text, ARIA use and landmark structure, on the same 0-100 bands.

SEO & best practices

Reports the Lighthouse SEO and Best Practices scores from the same run: crawlability, meta tags and structured data on one side, HTTPS, console errors and deprecated APIs on the other, each 0-100.

A free Lighthouse+ AI audit runs Google PageSpeed Insights on mobile and returns 4 Lighthouse category scores. Google bands each score, Performance, Accessibility, Best Practices and SEO, as good (90-100), needs improvement (50-89) or poor (0-49). UpMonitor runs this audit free with no account; PRO and Agency run the same checker on monitors, the CLI, the API and MCP. Every Active Incident AI diagnosis runs a fresh audit on the degraded monitor. The page returns a cached result, if the URL ran in the last 5 minutes.

What Is a Lighthouse Audit?

A Lighthouse audit is a Google PageSpeed Insights run that scores one page from 0 to 100 in 4 categories: Performance, Accessibility, Best Practices and SEO. The scores come from lab data. PageSpeed Insights loads the page once on a simulated mid-tier phone on a throttled mobile network, so the result is one controlled load and not field traffic. Every category uses the same 3 bands: 90-100 is good, 50-89 needs improvement and 0-49 is poor. A Lighthouse audit scores a page; it does not watch a server. It sits next to uptime, HTTP and response-time checks in a monitoring workflow and does not replace them. Lighthouse+ uses the mobile strategy only, so there is no desktop run.

What Does Lighthouse+ Add to a Lighthouse Audit?

Lighthouse+ adds 3 things to the raw PageSpeed run: a monitor checker that runs on a schedule, machine access through the CLI, the API and MCP, and a feed into the Active Incident AI diagnosis. The monitor checker runs inline with the monitor's other checkers at the plan interval, so there is no separate Lighthouse schedule. Machine access uses one checker name everywhere. The CLI takes --checks lighthouse, the API takes the same name on /v1/run-check, and the MCP server exposes run_audit with a URL and an optional checkers list.

The third addition is the one a plain PageSpeed run lacks. When a monitor degrades, the diagnostic worker runs a fresh mobile audit with a 30-second timeout. It keeps up to 6 headline metrics and the top 5 opportunities, sorted by potentialSavingsMs, the estimated milliseconds each fix saves. Gemini reads those numbers next to the failing checker output and writes 1 paragraph. The docs page on how the Active Incident AI diagnosis reads the Lighthouse run documents the diagnosis panel, its 24-hour cache and its monthly quota. The diagnosis runs without the audit, if PageSpeed Insights fails or exceeds 30 seconds.

How to Run a Free Lighthouse+ AI Audit in 4 Steps

The free audit takes 4 steps and no account: enter the URL, run the audit, read the 4 scores, and repeat the run from a terminal. The page sends one checker, lighthouse, to the public audit endpoint, so no other check runs and no plan gate applies. Steps 1 to 3 work for every visitor. Step 4 needs a PRO or Agency API key. The 4 steps run in this order:

  1. Enter the full URL with its scheme, for example https://upmonitor.io. The page must be public. PageSpeed Insights fetches the page from Google's servers, so a login wall, an IP allowlist or a firewall rule that blocks Google's servers fails the run.
  2. Run the audit and wait for PageSpeed Insights to finish. Google loads the page on the simulated phone and scores all 4 categories in one pass. The wait depends on Google and on your page. For scale, the Active Incident runner allows the same call 30 seconds.
  3. Read the badge, then open Details for the 4 scores. The badge reads Pass, Warning or Fail. The Details section prints the 4 scores as decimals from 0 to 1, for example "accessibility": 0.94. Multiply each decimal by 100 to place it on Google's 0-100 bands.
  4. Repeat the run from a terminal on a PRO or Agency key. The CLI check command options include --checks, which runs the Lighthouse+ checker alone. A free key gets HTTP 402 and no result.

Step 4 uses this 1 command:

bash
npx @upmonitor/cli check https://upmonitor.io --checks lighthouse

Verify a fix by running the audit again after the cache clears. The site caps anonymous runs at 5 audits per hour per IP address across all free tool pages and returns the cached result for 5 minutes per URL. A fix deployed 2 minutes after a run shows up only after the 5-minute cache expires. The page refuses the 6th run inside the same hour.

How to Read Your Lighthouse+ Results

The result is 1 badge and 4 scores, read against 3 bands. The 3 tiles above this article name what the audit checks: Performance score, Accessibility, and SEO and best practices. Each section below gives what the number means, the healthy line and the first fix step. The 3 bands share 2 lines, 50 and 90, the same lines Google uses.

Performance score

The Performance score is the weighted sum of 5 lab metrics; 90-100 is good, 50-89 needs improvement and 0-49 is poor. It is the only score that sets the checker status. The checker reports success at 50 and above and warning below 50. A low score never produces failure, so a slow page cannot mark a monitor down. The message line reads Audit complete. Perf: 42, SEO: 91 and shows 2 of the 4 scores. A healthy page holds 90 or more. First fix step: open the full PageSpeed Insights report for the URL and start with the opportunity that saves the most milliseconds.

Accessibility

The Accessibility score is the weighted share of Lighthouse accessibility audits the page passes, on the same 0-100 bands. The audits cover 5 areas: color contrast, form labels, image alt text, ARIA use and landmark structure. A score of 90 or more is good. The score does not change the checker status, so a page with a poor Accessibility score still shows Pass when Performance holds 50 or more. An automated audit tests what a machine detects. It does not replace a keyboard test or a screen reader test. First fix step: fix text contrast and missing form labels, because one CSS or template change repairs every page that shares the template.

SEO and best practices

The SEO score checks crawlability and on-page basics; the Best Practices score checks HTTPS, console errors and deprecated APIs. Both scores come from the same run and use the same 3 bands. The SEO category covers crawlability, meta tags and link basics. An SEO score of 100 means the page passes Lighthouse's basic checks. It does not measure rankings or content quality. Neither score changes the checker status, so read both in the Details section after every run. First fix step: let crawlers reach the page, write a valid meta description, serve every asset over HTTPS and clear the errors in the browser console.

The 5 Metrics Behind the Performance Score

Lighthouse, from version 10 on, weights 5 metrics into the Performance score: Total Blocking Time 30 percent, Largest Contentful Paint 25 percent, Cumulative Layout Shift 25 percent, First Contentful Paint 10 percent and Speed Index 10 percent. Lighthouse maps each metric onto a 0-100 curve, then takes the weighted sum. The weights explain a low score that the score alone hides. The table lists the 5 metrics from the heaviest weight to the lightest, with the good line for each on mobile.

MetricWeightWhat it measuresGood on mobile
Total Blocking Time30 percentMain thread time blocked by long tasks after the first paint200 ms or less
Largest Contentful Paint25 percentTime until the largest image or text block is painted2.5 s or less
Cumulative Layout Shift25 percentHow far visible elements move while the page loads0.1 or less
First Contentful Paint10 percentTime until the first text or image is painted1.8 s or less
Speed Index10 percentHow fast the visible part of the page fills in3.4 s or less

Total Blocking Time and Largest Contentful Paint hold 55 percent of the score together, so JavaScript weight and the largest image decide most of the result. The Active Incident runner keeps 6 headline metrics. 5 of the 6 are the weighted metrics in this table. The 6th is server-response-time, the time the server takes to send the first byte.

Five amber and brown bars of falling height on one dark plate, one bar per weighted Lighthouse metric

Is Lighthouse+ Free?

The one-off audit on this page is free and needs no account; the Lighthouse+ checker on monitors, the CLI, the API and MCP needs the PRO or Agency plan. Lighthouse+ is the 1 paid checker in the checker registry. The API answers a free key with HTTP 402 and the message The Lighthouse checker requires a PRO or Agency tier. The dashboard disables the Lighthouse+ toggle on free monitors. The public page has no plan gate, only the limit of 5 runs per hour per IP address. The 3 plans differ on 4 numbers, and the table shows each one.

PlanPrice per monthShortest intervalRegionsAI diagnoses per month
Free$0300 seconds15
PRO$960 seconds650
Agency$3260 seconds12Unlimited

The PRO and Agency plan limits page lists every other limit. On both paid plans the Lighthouse+ checker runs at the monitor interval, with no separate Lighthouse schedule to set. The free audit on this page stays free after an upgrade, with the same 5-minute cache.

Common Lighthouse+ Issues and How to Fix Them

Lighthouse+ runs hit 5 known issues: 2 checker failures, 1 low-score warning, 1 plan error and 1 score mismatch. Each item gives the symptom, the cause, the fix and the prevention. 4 of the 5 symptoms are exact strings from the checker and the API, so they match what you see on screen. The 5 issues are:

  1. Lighthouse audit failed: PSI API responded with 500 (also 429 or 400). Cause: PageSpeed Insights returned an error status and no report. Fix: Ensure the URL is publicly accessible and not blocking Googlebot. Confirm the URL returns 200 without a login, check firewall and WAF rules for a block on Google's servers, then run again after 5 minutes. Prevention: a monitor with the Lighthouse+ checker reports the same failure at its next interval, so a new firewall rule shows up within 1 interval, 60 seconds at the shortest.
  2. Audit complete. Perf: 42, SEO: 91 with a Warning badge. Cause: the Performance score is below 50. Fix: Review the detailed Lighthouse report to identify performance bottlenecks. Then size images to their render box in WebP or AVIF, defer render-blocking CSS and JavaScript, and preload the Largest Contentful Paint asset. Prevention: a PRO monitor turns the 50 line into a recurring warning.
  3. HTTP 402 with The Lighthouse checker requires a PRO or Agency tier. Cause: a free key asked for the paid checker from the CLI, the API or MCP. Fix: run the audit on this page, or upgrade the plan. Prevention: none needed, because the public page stays free.
  4. Two runs return different scores, or the score differs from Chrome DevTools. Cause: lab data varies between loads, a desktop DevTools run is not a mobile run, and this page caches results for 5 minutes. Fix: compare mobile to mobile, wait out the cache, and judge by band and not by single points. Prevention: a monitor records the score at every interval, so one odd run stands out.
  5. Lighthouse audit skipped: PSI_API_KEY not configured. with the error code MISSING_API_KEY. Cause: the worker that ran the check had no PageSpeed Insights key, so nothing on your site is wrong. Fix: run the audit again, and report the message if it repeats. Prevention: a monitor repeats the check at its next interval, so 1 skipped run does not hide the next score.

How Lighthouse+ Fits Continuous Monitoring

On a monitor, the Lighthouse+ checker runs inline with the monitor's other checkers at the plan interval, and its warning state counts toward the monitor's health. The checker's impact ceiling is warning, so a low score turns health amber and never red. When health degrades, the diagnostic worker runs its own fresh PageSpeed audit. It passes the 4 scores, up to 6 headline metrics and up to 5 opportunities to Gemini with the failing checker output. Gemini returns 1 paragraph, cached for 24 hours per incident signature. Each monitor gets at most 1 diagnosis every 15 minutes. The monthly quota is 5 diagnoses on Free, 50 on PRO and unlimited on Agency. A PRO or Agency monitor with the Lighthouse+ checker applies the 50 line at every interval.

The 3 free checks below pair with a Lighthouse score, and each one isolates a number the score folds in. Server-response-time is 1 of the 6 headline metrics, so the free website performance test isolates time to first byte. A heavy HTML document delays First Contentful Paint, so the free page weight checker sizes the initial HTML and flags anything past 500 KB. The Best Practices score flags HTTPS problems, and the free security headers check explains the response headers one by one. All 3 run on this site with no account, the same as this audit.

Also useful for this stack

Pair the audit you just ran with these checkers - the failure modes tend to travel together.

Stop running this audit by hand.

Schedule Lighthouse+ AI Bridge every 5 minutes free, or unlock 1-minute sweeps across all 12 regions with PRO. Get an AI-drafted remediation prompt the moment a check fails - delivered to your inbox, Slack, or MCP-connected agent.