SEO Health
SEO Health answers one question for a site owner: can search engines find, index and show my important pages well, and what should I fix first? It is the search side of a site, beside the AI side that the rest of Vercatus measures. This page is everything behind its number.
The idea in five lines
Section titled “The idea in five lines”- Facts, not opinions. Every result comes from something measured on a page: a status code, a tag, a length, a timing. No AI model reads a page or produces the score.
- A rulebook with sources. Each rule is a documented search-engine requirement (Google Search Central, Bing, web.dev, Lighthouse). Where no engine documents a rule, it is established practice, is marked as such and weighs low.
- Important pages weigh more. A problem on the home page costs three times what it costs on an ordinary page; on a page linked from home, cited by AI answers or earning search clicks, twice.
- Failures subtract. Each question starts at 100 and loses points for each problem, so a site cannot hide a serious fault behind many small passes.
- Every point is accounted for. The points lost add up to 100 minus the score, and each lost point belongs to a named rule with the pages it was found on.
Where the data comes from
Section titled “Where the data comes from”SEO Health runs its own light check of the site, apart from a full run: once a week for a verified site, and on Check again (at most once an hour). Until a site’s first check finishes, the tab reads the pages of its latest run.
| Read by the check | What the rules read from it |
|---|---|
| Each sampled page, fetched once as an ordinary client, politely | Status and redirects, response headers, and the HTML as written: title, description, robots rules, canonical, hreflang, headings, structured data, images, links |
| A few pages in a headless browser | Lab speed, the main image’s loading, the width at a phone’s width |
| A few extra requests | A made-up address (soft 404), the http:// and www forms of the site, robots.txt and the sitemaps, the link and canonical targets the sample did not read |
| Real-visitor speed data from Google, where the site has enough traffic | Field LCP, CLS, INP and TTFB, preferred over lab |
| Search Console and the answers Vercatus collects, when connected | Which pages matter: clicks in the last 28 days, citations in the last 90 |
Which pages are read
Section titled “Which pages are read”The check reads a sample: the home page, the key pages linked from it, the pages with traffic or AI citations, the pages the last check read (so checks compare like with like), a fair spread of every template, and newly listed pages. The sample size is set by the plan: 50 pages a site on Starter, 150 on Growth, 300 on Portfolio, 500 on Enterprise.
A large site is never crawled in full. Most SEO faults are template faults, so a few pages of each template stand for all of it, and the tab shows the share found on each template (“27 of 32 knowledge pages”). It always says what it read.
| Page | Weight |
|---|---|
| Home page | 3 |
| Key page, cited page, visited page | 2 |
| Any other page | 1 |
The four questions
Section titled “The four questions”Rules are grouped by what they mean to the owner, not by technical category. They weigh differently because a page that cannot be indexed gains nothing from a good title.
| Question | Weight in the site score |
|---|---|
| Can search engines find and index it? | 35 |
| How does it look in search results? | 25 |
| Is the content well built? | 20 |
| Is it fast and safe for visitors? | 20 |
How a rule costs points
Section titled “How a rule costs points”Each rule reads a page, or the site, and answers pass, partial (half credit) or fail, with what it measured (“64 characters”, “3,200 ms”). A rule that does not apply (language rules on a single-language site) or could not be measured gives no answer: it costs nothing and is listed as not checked.
- A rule’s share. The share of the pages it applies to that pass, each page counted by its weight.
- A question’s score. It starts at 100. Each rule loses its severity’s penalty times the share that does not pass: critical 60, high 30, medium 15, low 5. Related rules form a family, and a family costs its worst rule’s loss, not the sum. The low rules of a question together cost at most 15.
- The site score. The four scores, weighted 35, 25, 20, 20.
- Caps. A fault that sinks the site holds the score down whatever else passes: a home page that errors or is set to noindex, robots.txt blocking Googlebot, most checked pages unindexable, or no HTTPS.
- Never 100 with something open. While any problem remains, the score is 99 at most.
90 to 100 is good, 70 to 89 needs work, and below 70 is poor.
What SEO Health does not do
Section titled “What SEO Health does not do”- No backlinks, keyword research or rank tracking. That needs search-index data Vercatus does not hold. Topics and Plays cover demand.
- No AI judgement in the score. Nothing a model writes changes how the score is made.
- No full crawl of large sites. It reads a sample and says what it read.
Limits to know
Section titled “Limits to know”- “No links in” counts only links among the checked pages, so it is low severity and says so.
- Link and canonical targets are requested in a small number beyond the sample, the most linked first. A link to a page that robots.txt closes to us, or on another site, is not followed.
- Speed and browser rules read a handful of lab pages, and field data where Google has it; the other pages are not checked there.
- The content width is measured in a desktop browser narrowed to a phone’s width, not on a phone.
A problem can be sent to Plays, where the next check verifies the fix. The same score is available through the API. To read the tab itself, see Reading SEO Health.