Origin of the term
When Google added the SEO category on 5 February 2018, its developer blog explained the name: the tool "keeps you from crashing into the rocks", hence Lighthouse.
Glossary Google Lighthouse
Google Lighthouse is an open-source tool from Google that audits a web page under lab conditions. It loads the URL on a simulated device and connection, runs a set of checks and returns a report with a score from 0 to 100 for each category: performance, accessibility, best practices and SEO. The numbers come from one test run, not from real visitors; page speed explains the difference between lab and field data.

Google Lighthouse: Google's open-source tool that audits a URL in a lab run. Categories, how the performance score is weighted, modes, limits.

When Google added the SEO category on 5 February 2018, its developer blog explained the name: the tool "keeps you from crashing into the rocks", hence Lighthouse.
The public Lighthouse repository on GitHub dates from 8 March 2016. The milestones below come from Google's blog and the project's release notes:
Lighthouse opens the page in Chrome, loads it and records what happens. By default it does not slow the network down for real. It runs a fast load and then uses a model to estimate how the same load would have gone on a slow connection. The default network profile, "Slow 4G", assumes 150 ms of latency and 1.6 Mbps download, according to the project's throttling documentation. The docs describe this simulated throttling as fast and deterministic, with some inherent inaccuracy.
From that data Lighthouse runs the audits for each category. Only five metrics feed the performance score. The opportunities and diagnostics in the report help you improve those metrics, but they add nothing to the score themselves. Each raw metric value is turned into a score from 0 to 100 using a log-normal curve that Google derives from real-site data in the HTTP Archive. Lighthouse then takes a weighted average.
Performance score weights in Lighthouse 13.6.0:
| Metric | Weight |
|---|---|
| Total Blocking Time (TBT) | 30 % |
| Largest Contentful Paint (LCP) | 25 % |
| Cumulative Layout Shift (CLS) | 25 % |
| First Contentful Paint (FCP) | 10 % |
| Speed Index (SI) | 10 % |
Source: Lighthouse 13.6.0 default configuration on GitHub and Chrome for Developers, "Lighthouse performance scoring", accessed 17.11.2026.
Google's scoring page says "Last updated 2019-09-19", yet its table is headed "Lighthouse 10", a version from 2023. The release notes for 10.0 link to that same table, and the source code of 13.6.0 still holds the same values. So the weights are the version 10 weights, unchanged since February 2023; the page stamp is simply older than the table.
Three of the five metrics have no entry of their own in this glossary. First Contentful Paint marks when the first content appears. Speed Index measures how quickly the visible area fills in. Total Blocking Time adds up the time after the first content during which the main thread is busy enough that the page would not respond to input. LCP and CLS have their own entries.
INP, the responsiveness metric in Core Web Vitals, is not part of the score. A navigation run only watches the page load, and nobody clicks anything, so INP cannot be measured in that mode. Lighthouse measures TBT instead. web.dev says TBT "may be a reasonable proxy metric for INP for the lab" and adds that it is "not a substitute". Only timespan mode records INP, and only for the interactions you perform during the run.
The other categories work differently. Accessibility is also a weighted average, with weights based on axe user-impact assessments; each audit either passes or fails. Since version 13.3 there is also the Agentic Browsing category, which Google labels experimental. It shows how many of its checks the page passes rather than a 0 to 100 score.
Lighthouse has three modes, according to its GitHub documentation:
| Mode | What it analyses | What it does not give you |
|---|---|---|
| Navigation | A single page load; the default mode | Form submissions or content that appears after the load |
| Timespan | A period of time, usually with interactions | An overall performance score or LCP |
| Snapshot | The page in one particular state, such as an open menu | Performance metrics |
Source: GitHub, GoogleChrome/lighthouse, "User Flows in Lighthouse", accessed 17.11.2026.
You can run it in four ways, as listed in Chrome for Developers' introduction:
| Route | Who it suits | Note |
|---|---|---|
| Chrome DevTools | Development and manual review | Works on pages behind a login |
| Command line | Scripted, repeatable runs | Needs Chrome installed |
| Node module | Continuous integration, user flows | Can combine several steps into one report |
| PageSpeed Insights | Anyone, no installation | Also shows field data |
Source: Chrome for Developers, "Introduction to Lighthouse", accessed 17.11.2026.
The Chrome extension that Google used for the SEO launch in 2018 is not among the routes in today's introduction.
A worked example with round numbers. A page gets these metric scores: TBT 50, LCP 80, CLS 100, FCP 90 and Speed Index 70. The performance score is:
50 × 0.30 + 80 × 0.25 + 100 × 0.25 + 90 × 0.10 + 70 × 0.10 = 15 + 20 + 25 + 9 + 7 = 76
Now the team cuts main-thread blocking and TBT goes from 50 to 80. The score rises by 9 points to 85. Improving FCP by the same 30 points would add only 3.
A second case shows the limits of the number. A shop runs Lighthouse once on a fast laptop and gets 92, then runs PageSpeed Insights and gets 68. Both can be right. They measured on different hardware, and PageSpeed Insights applies its own simulated device. Comparing the two numbers says nothing about whether the site changed.
Lighthouse colours every score in one of three bands:
| Score | Colour | Meaning |
|---|---|---|
| 0–49 | Red | Poor |
| 50–89 | Orange | Needs improvement |
| 90–100 | Green | Good |
Source: Chrome for Developers, "Lighthouse performance scoring", accessed 17.11.2026.
Google calls a perfect 100 "extremely challenging to achieve and not expected". The curve flattens at the top: going from 99 to 100 takes about as much metric improvement as going from 90 to 94.
Scores also move between runs when nothing in the code has changed. The project's variability guide lists page nondeterminism such as A/B tests and ads, local network variability, client hardware and other processes competing for resources. Its advice is to run Lighthouse several times and use the median, the 90th percentile or min and max rather than a single result. In the project's words, "the median Lighthouse score of 5 runs is twice as stable as 1 run".
| Use case | For | Against |
|---|---|---|
| Testing before a release | Works on staging and on pages behind a login | The device is simulated and may not match your visitors |
| Comparing two versions | Same conditions, diagnosis audit by audit | Without several runs, the difference may be noise |
| Accessibility review | Finds common failures in seconds | Detects only a subset of issues |
| Monitoring in CI | Scriptable through the Node module | Needs stable test hardware to produce comparable numbers |
Source: own assessment based on the Lighthouse documentation, accessed 17.11.2026.
Lighthouse is the lab engine. What real visitors experience shows up in field data, which page speed and Core Web Vitals explain.
| Tool | Data type | Where the data comes from | Use it for |
|---|---|---|---|
| Lighthouse | Lab | One run on a simulated device | Diagnosing issues, comparing versions |
| PageSpeed Insights | Lab and field | Lighthouse plus CrUX | Seeing both on one page |
| CrUX | Field | Chrome users, trailing 28 days | Knowing how the site performs for people |
| Search Console Core Web Vitals report | Field | CrUX, grouped by similar URLs | Finding groups of pages with problems |
Source: PageSpeed Insights documentation and Search Console Help, accessed 17.11.2026.
The SEO category does not measure rankings. Google said so in 2018: its SEO audits make no guarantees for web search. On its page experience documentation (updated 22.09.2026), Google states that Core Web Vitals are used by its ranking systems. It adds that good results in Search Console or third-party tools do not guarantee top rankings, and that chasing a perfect score just for SEO "may not be the best use of your time". The Core Web Vitals in Search Console come from real users, not from a Lighthouse run. Google does not name the Lighthouse score as a signal on the pages read for this entry.
On accessibility, Lighthouse warns about itself. The category description in version 13.6.0 says automatic detection "can only detect a subset of issues and does not guarantee the accessibility of your web app". A score of 100 does not show conformance with WCAG or the European Accessibility Act. Manual testing is still needed.
This entry was written in November 2026 with AI assistance. The facts come from Chrome for Developers, web.dev, Google Search Central, Search Console Help and the Lighthouse repository on GitHub, all accessed on 17.11.2026. The worked example is our own calculation. The entry was reviewed before publication and contains no client data.
A technical audit covers this and everything else in one pass.
Request an auditYes. Lighthouse is open source under the Apache 2.0 licence. It is built into Chrome DevTools, can be installed as a Node module, and also runs inside PageSpeed Insights, where you need to install nothing at all.
Google marks 90 to 100 green, 50 to 89 orange and anything below 50 red. It says a perfect 100 is extremely challenging and not expected. Look at the median of several runs rather than one result.
Lighthouse runs the lab test. PageSpeed Insights runs that same test in a simulated environment and adds field data from real Chrome users over the previous 28 days, when there is enough of it.
Because of changes on the page such as ads or A/B tests, network conditions, hardware or other processes running. The Lighthouse team recommends several runs and the median; five runs are twice as stable as one.
Google says its ranking systems use Core Web Vitals, measured with real users, and that a perfect score does not guarantee rankings. It does not name the Lighthouse score as a ranking signal in its documentation.