Skip to content
search-engine-optimizedby Netsolution

Lighthouse

It is an open audit tool, published by the browser's vendor, which loads a page under conditions it sets and hands back scores per panel. It assesses one page at a time, in the lab, and replaces neither a crawler nor a measurement of real audience.

What it does

It loads a page under conditions it sets — simulated device, throttled network — then hands back a series of findings sorted into panels: loading, accessibility, best practices, and a panel named “SEO” that checks a list of technical points. Every finding comes with the offending element, which makes it a diagnostic tool more than a measuring one.

The tool is published under the Apache 2.0 licence, declared in the public repository hosting its code.

officialSource : Lighthouse, public repository and declared licence

The repository belongs to the browser's own organisation: the tool is published by the vendor, and not by a third party or an independent community.

officialSource : Lighthouse, public repository of the browser's organisation

That second claim is what makes this entry particular, and it is worth stating plainly: the browser’s vendor publishes the tool that scores the pages, under an open licence. The code is checkable by anyone, which is a real advantage over tools whose formula stays private. But an open code base does not make the weighting neutral: it makes it legible. Those are two different things, and the second is already a great deal.

What it does not do

An audit loading a page under conditions fixed in advance produces a lab measurement: it is reproducible and comparable with itself, but it describes no real visitor's experience.

hypothesis

It does not crawl a site: it examines one address at a time. Yet a site’s costliest faults are faults of distribution — pages returning an error, repeated titles, redirect chains — and they only appear in a complete inventory.

Nor does it say what the engine kept of your site, which is read in its console. And it does not measure what your visitors actually experience, which is field data.

An audit score is calculated by the tool by its own weighting: no engine publishes a per-page score used for ranking, and a panel named “SEO” inside an audit remains a list of technical checks, not a forecast of position.

hypothesis

The panel named “SEO” deserves a warning of its own: it checks that a page respects a list of elementary controls — title present, page crawlable, text legible. Passing it in full means no gross error was made. It has never meant the page would rank, and the panel’s name alone sustains that confusion. The subject is handled in detail in the guide on speed and loading experience.

What it costs

Nothing. The tool is free, open, and built into the browser’s developer tools: there is no compulsory installation, no account to create, and no paid hosted offer from the vendor recorded to date. Of every entry in this directory, it is the one whose ratio of usefulness to entry cost is the most favourable — and that is precisely why it has to be clear about what its score does not say.

When a paid tool stays useful

As soon as the question moves from “why is this page slow?” to “how many pages on the site are affected?”. Auditing one page says nothing about the scale of a problem, and that is exactly what a crawler — Screaming Frog SEO Spider in this section — can do and it cannot.

The two complement each other better than they replace each other: the crawler finds where to look, the audit explains what to fix. Starting with the audit on a page picked at random often means optimising at length a page whose problem had no scale at all.

Auditing a page with the browser's tool before buying anything

hypothesis

Effort
low
Time to effect
immediate
Risk
none
Durability
structural

Affiliation

This site takes part in no affiliate programme for this tool, and none exists: an open tool published by a browser’s vendor pays nobody to be recommended. This entry earns nothing, and the order of this section has never depended on payment.

What does not work

Get your Lighthouse score green on every panel and Google will reward you in its results.

The score is produced by the tool by a weighting it publishes, and not by the engine by a ranking scale — no engine publishes a per-page score that position depends on. The measurement is moreover made in the lab: it varies from run to run and describes no real visitor's experience. A perfect score establishes that no gross error was made on that page, and nothing more.

Sources

  1. Lighthouse — public repository and licence — Googleaccessed on
  2. About Search Console — Google Search Consoleaccessed on

Block revalidated on

Entries in this category

Frequently asked questions

Does a good score guarantee a better ranking?
No, and that is the costliest confusion about it. The score is calculated by the tool by its own weighting, which it documents. No engine publishes a per-page score that ranking depends on, and improving the score has never amounted to a promise of position.
Why does my score change on every run?
Because an audit really loads the page, with the machine, the connection and the load of the moment. The measurement is therefore noisy by nature. You read a trend over several runs, never a few points of difference between two consecutive readings.
What is the difference from field data?
The audit simulates a load under conditions it sets; field data aggregates loads actually experienced by visitors. The first explains why a page is slow, the second establishes whether it is slow for your visitors. They are read together.
Does it replace a site crawler?
No, it works page by page where a crawler works on the complete inventory. The faults that count most are often faults of distribution, visible only in a sorted list of every address. The two tools answer different questions.
Does it have to be installed?
Not necessarily, since it is built into the browser's developer tools, which is amply enough for auditing a page occasionally. Installing it separately only matters for automating it, wiring it into a publication pipeline or keeping a history of successive readings over time.