Lighthouse SEO: the ten checks behind the score, and why one of them outweighs the rest

Lighthouse SEO is the SEO category inside Google's Lighthouse tool, the auditor built into Chrome DevTools and PageSpeed Insights. It runs ten automatic checks on one page, such as whether the page can be indexed, whether it has a title and a meta description, and whether its links can be crawled, then gives a score from 0 to 100. One check, whether the page is blocked from indexing, carries about four times the weight of the others, so failing it alone pulls the score down to around 69. A high score means the page passes these basic hygiene checks. It is a starting floor for technical SEO, and the rest of ranking is decided by content, links and relevance.

How to run a Lighthouse SEO audit

There are three common ways, and they run the same audits.

• Chrome DevTools. Open the page in Chrome, press F12, choose the Lighthouse panel, tick SEO, and click Analyze page load.
• PageSpeed Insights. Paste a URL at pagespeed.web.dev. The SEO score sits beside Performance, Accessibility and Best Practices in the lab section.
• The command line. With Node installed, npx lighthouse https://example.com --only-categories=seo runs the SEO category alone, and --output=json gives a file you can store and compare month to month.

The score is coloured the same way as the other categories: 90 to 100 is green, 50 to 89 is orange, and 0 to 49 is red.

The ten automatic checks, and what each one wants

We read the list from the Lighthouse source code on GitHub on 21 October 2026. Lighthouse groups the checks into two sets: whether search engines can crawl and index the page, and whether they can understand its content.

Audit	Group	What fails it
Page is blocked from indexing	Crawl and index	A noindex directive in a meta robots tag or an X-Robots-Tag header, or a robots.txt rule that blocks the page
Page has unsuccessful HTTP status code	Crawl and index	The page returns a 4xx or 5xx status
Links are not crawlable	Crawl and index	Anchors with no usable href, such as links that only work through a JavaScript click handler
robots.txt is not valid	Crawl and index	Syntax errors or unknown directives in the robots.txt file
Document does not have a title element	Content	No title tag, or an empty one
Document does not have a meta description	Content	No meta description, or an empty one
Links do not have descriptive text	Content	Anchor text such as "click here", "read more" or "learn more"
Image elements do not have alt attributes	Content	Images with no alt attribute
Document does not have a valid hreflang	Content	hreflang values with an invalid language or region code
Document does not have a valid rel=canonical	Content	Conflicting canonicals, an invalid URL, or a canonical that points somewhere it should not

An eleventh item, "Structured data is valid", is listed as a manual check. It carries no weight in the score, and Lighthouse asks you to test the markup yourself with Google's Rich Results Test or the Schema Markup Validator.

How the score is weighted

Nine of the ten automatic audits carry a weight of 1. The indexing check carries 93/23, roughly 4.04. A comment in the Lighthouse code explains why: that one audit is meant to be at least 31 percent of the category, so a page that tells search engines not to index it fails the SEO category on its own.

The arithmetic is simple. The total weight is about 13.04. Fail only the indexing check and you keep about 9 out of 13.04, a score near 69. Fail one ordinary check, such as a missing meta description, and you keep about 12.04 out of 13.04, a score near 92. Audits that do not apply to a page are left out of the sum, so two pages with different templates can reach the same score through different sets of passes.

This weighting tells you where to look first. A score in the 60s almost always means a noindex or a robots.txt block. A score in the low 90s usually means one small content item.

Fixing the common failures

Blocked from indexing

Check three places: the page source for a meta robots tag, the response headers for X-Robots-Tag, and robots.txt for a Disallow rule covering the path. The usual cause is a staging setting copied to production, or a CMS option such as "discourage search engines" left switched on. If the block is deliberate, for a thank you page or an internal search page, the failing score on that page is correct and you can ignore it.

Links are not crawlable

Every link that should be followed needs an anchor element with an href holding a real URL. Buttons that navigate through script, and links with href set to "#" or "javascript:void(0)", fail. This matters most on navigation menus and pagination in single page apps.

Links do not have descriptive text

Rewrite the anchor so it says where it goes: "see the pricing table" rather than "click here". This helps screen reader users as much as search engines.

Missing title or meta description

Both belong in the page template, filled from fields your CMS already has. Lighthouse checks only that they exist and are not empty, so a passing audit says nothing about whether the title is a good one.

Invalid hreflang or canonical

Use valid ISO language codes, such as en, en-GB or es-MX, and absolute URLs. Each page should carry one canonical that points at the version you want indexed.

What a perfect Lighthouse SEO score covers, and what sits beyond it

A score of 100 confirms the page clears the technical floor on that one URL, at that moment. Ranking draws on much more: whether the content answers the query, whether other sites link to it, how fast it loads for real visitors, and how it compares with the other pages competing for the same search.

So treat Lighthouse SEO as one line in a wider check. Speed has its own category, and our guide to reading the PageSpeed Insights report covers the field data that Lighthouse does not see. For the site level items a single page audit cannot reach, such as sitemaps, redirects, duplicate pages and internal linking, the technical SEO checklist walks through every check in order.

One habit makes Lighthouse far more useful: run it on each page template rather than only the homepage. A product page, a category page, a blog post and a location page each come from different code, and one broken template can affect thousands of URLs.

The newer audit category beside the SEO one

The same Lighthouse source we read for this page also carries a separate category called Agentic browsing, with checks such as whether a site publishes an llms.txt file. It is a sign of where the toolmakers think the next audit sits: not only whether a search crawler can index a page, but whether an AI assistant can read and use it. A page can score 100 on SEO and still be absent when somebody asks ChatGPT or Gemini which business to choose. That second question is measured by asking the assistants directly, and AI Knows Us is built to do exactly that, showing which businesses each one names and which pages it read to decide.

Questions about Lighthouse SEO

Does a 100 Lighthouse SEO score improve my Google rankings?

Not by itself. Google does not use the Lighthouse score as a ranking signal. The audits flag problems that can stop a page being crawled or understood, and fixing those removes a barrier.

Why is my Lighthouse SEO score 69?

That number almost always means the page is blocked from indexing, because that single audit carries about four times the weight of the others. Check for a noindex tag, an X-Robots-Tag header or a robots.txt rule.

Why do I get a different SEO score in DevTools and PageSpeed Insights?

The two can run different Lighthouse versions, and extensions in your own Chrome profile can change the page. Run DevTools in an incognito window for the cleanest comparison.

Does Lighthouse check structured data?

Only as a manual item with no weight. Test markup with the Rich Results Test or the Schema Markup Validator.

How often should I run Lighthouse SEO?

After every template or CMS change, and monthly on one page of each template. Saving the JSON output lets you see exactly which audit changed.