What the health score means

The score starts at 100 and loses a weighted penalty for each open finding. The weight is how much money the fault stands in front of, which is not the same as how broken it is: a missing product image costs more than a long meta description even though neither is an error.

The four categories

Technical. Canonicals, redirects, robots directives, sitemaps, broken links, indexability.

Content. Titles, descriptions, headings, thin pages, duplicates.

Speed. Response time and page caching.

WooCommerce. Product schema, variations, faceted URLs, stock states, category pages.

Each category is capped at 25 points. That cap is doing real work: a shop with 1,800 products missing alt text cannot lose its whole score to one fault and hide everything else. Without it, one systematic problem saturates the number and every other problem becomes invisible.

Why the penalty is not proportional

The penalty grows with the square root of the affected pages rather than in proportion to them.

The reasoning is that the second instance of a fault tells you less than the first. Forty pages with no meta description is worse than one page with none, but it is not forty times worse, because it is almost always one template and one fix. Scoring it as forty times worse would make the number swing wildly with the size of the site rather than with its quality.

What a good score is

80 and above is good. 70 to 79 is fair. Below 70 needs work.

Below 70 on a site that has never been audited is completely normal, and it is usually three or four systematic causes rather than two hundred individual ones.

Our own site scored 52 the first time we pointed the crawler at it.

Muting

A muted finding does not count. Muting is you saying it is intentional.

Use it. A contact page with 40 words on it is thin content by measurement and complete by intent, and a finding you have read and dismissed has done its job. The mute records the reason so the next person to look does not undo it.

What the score cannot tell you

Whether your content is any good. No crawler knows that.

Whether you will rank. Health is hygiene. A perfectly healthy site with nothing to say about a topic will not outrank a messy one that is the best answer to the question.

That every finding is real. This is worth saying about our own product as much as anyone else's. In September 2026 we read which addresses our own findings pointed at and found that most of them were not problems: markup rules running against JSON files, a decorative logo with correct empty alt text reported on 37 pages, sign-in pages marked down for having no search description on pages that ask Google not to list them. All of that is fixed, in the crawler, in the hub and in all seven plugins, from one shared definition of how a page is graded.

A check that fires on correct markup teaches you to ignore every other check it makes, and that is a worse failure than missing something.

Audit read 2 hours ago, 1,204 pages Sample data

Site audit, down 3 since last month17 findings open across four categories. Two of them are worth doing first.

Technical6 Content5 Web Vitals2 WooCommerce4

• High 3 pages Canonical points to another page /pricing/?plan=agent-plus canonical is /pricing/
• High 2 pages Redirect chain of 3 hops /features/reporting to /client-reports to /reports/
• Medium 4 pages Duplicate title "SEO Tools Australia" on 4 documentation pages
• Low 1 page In sitemap, answers 404 /news/2025-beta/

Using it

The score sits above the findings it came from, so the number and the reason for it are on one screen. Sort by weight rather than by count, fix the systematic causes, and re-crawl. One template change usually closes dozens of findings at once and moves the category that was capped.

The arithmetic, worked

For each finding, the penalty is the rule's weight, multiplied by a factor for its severity, multiplied by the square root of how many pages carry it. The penalties in a category are added up and then capped at 25.

Take a site with 40 pages sending no cache header. That rule is weight 5 and sits in Speed. Forty pages is a square root of about 6.3, so the raw penalty is large, and the cap brings it to 25. Speed is now fully spent: a second speed problem on that site costs nothing more until the first one is fixed, which is exactly the signal you want.

Now take one page with a canonical pointing at a redirect. That rule sits in Technical, and one page is a square root of one, so the penalty is the weight alone. Small in the score, and often the most expensive thing on the list in reality, which is why the list is sorted by weight rather than by what it did to the number.

Why the number can fall after you fix something

Two ways, and both are honest.

A crawl that reads more pages finds more. Raising the page budget, or publishing forty products, means the crawler sees faults that were always there.

A fault that was being hidden becomes visible. If a category was capped, the findings underneath the cap were costing nothing. Fix the big one and the next one starts counting.

Neither is the score being unstable. It is the score being computed from what was actually read.