How the Site SEO AI Audit Score Is Calculated: Severity and Weights

A single number out of one hundred is the first thing you see in a Site SEO AI Audit report, and it is also the thing people argue about. Why did the SEO score drop from 78 to 71 when only two issues changed? Why does one critical error cost fewer points than a warning that appears on many pages? Why is the speed area worth more than the languages area? The answers are in the way the score is built, and once you know the rules, the number stops being a mystery and becomes a tool for deciding what to fix first. This guide explains the calculation as Site SEO AI Audit, a product of the Internet Solutions team, describes it: severity, share of pages, softening and area weights.
We use the figures published on the product’s own pages. Where we illustrate with an example, we label it as an illustration, and we do not claim to reproduce the exact formula beyond what the product states.
The score in one sentence
Each issue has a severity, and that severity is multiplied by the share of your indexable pages the issue affects, softened so that one forgotten page barely counts. The seven area scores are then combined by weight into the overall score. Three ideas are packed into that sentence: severity, share and weight.
Idea one: severity
Not every problem is equally serious. The audit classifies each issue into one of three levels and gives each a base value.
| Severity | Base value | What it usually means | Example from a sample report |
|---|---|---|---|
| Critical | 10 | Something that can keep pages out of search or confuse search engines about which page counts | A canonical address that points to a page that redirects |
| Warning | 4 | A problem that weakens pages but does not block them | Duplicate titles on many pages |
| Notice | 1 | A minor improvement or a hint | Not shown in the sample report |
Severity is the reason the report lists critical items first. A canonical tag is a signal that tells search engines which version of a page is the preferred one, and Google explains the idea in its guide to consolidating duplicate URLs. When that signal points to a page that itself redirects, search engines receive contradictory instructions, so the problem is classed as critical.
Idea two: the share of pages affected
The same issue means very different things on one page and on a third of your site. The audit therefore multiplies severity by the share of your indexable pages that the issue affects. Indexable means the pages that search engines are allowed to put in their results, so pages you deliberately keep out of search do not distort the figure.
Here is an illustration with the sample report published on the product page, which shows a site with 412 pages. The numbers below are only for understanding relative size, because the real calculation applies a softening step that we describe next.
| Illustrative issue | Severity value | Share of pages | Severity times share, before softening |
|---|---|---|---|
| Duplicate titles on 64 pages (warning) | 4 | about 15.5 percent | about 0.62 |
| A critical error on 3 pages | 10 | about 0.7 percent | about 0.07 |
| A notice on every page | 1 | 100 percent | 1.00 |
The lesson is simple and often surprising. A moderate problem that is widespread can matter more than a serious problem on a handful of pages, and a trivial problem repeated everywhere can outweigh both. That is why you cannot judge the report only by the colour of the labels.
Idea three: softening
The product states that the share is softened so that one forgotten page barely counts. This protects you from a common injustice: a large site with thousands of pages will almost always contain a few pages with some defect, such as an old page with no description. Without softening, those few pages would pull the score down just as much as a real structural fault. With softening, an issue on a tiny fraction of the pages has little effect, while an issue that touches a meaningful share grows quickly in importance. The exact shape of the curve is not published, and for decisions you do not need it. What matters is the principle: widespread issues count, isolated ones barely do.
Idea four: area weights
The score is built from seven area scores, each describing a different part of search visibility. The areas are combined by the following weights.
| Area | Weight | What it covers |
|---|---|---|
| Crawl and index | 25 percent | Status codes, robots.txt, noindex, canonicals, sitemap, redirect chains, orphan pages, click depth |
| On-page | 20 percent | Titles, descriptions, H1, thin and duplicate content, image alt texts |
| Links | 15 percent | Broken internal and external links, links to redirects, weakly linked pages, nofollow, empty anchors |
| Speed and Core Web Vitals | 15 percent | Google PageSpeed, LCP and CLS, server response, compression, page weight |
| Structured data | 10 percent | Schema.org on each page, invalid JSON-LD, Open Graph and share images |
| AI visibility | 8 percent | AI crawlers in robots.txt, llms.txt, content that needs JavaScript, structure and dates that AI answers quote |
| Languages | 7 percent | hreflang return links, broken language versions, x-default, lang attributes |
The weights add up to one hundred percent. They express a judgment about importance: if search engines cannot crawl and index your pages, nothing else matters, which is why that area carries the largest share. Structured data helps search engines understand your content, as Google explains in its introduction to structured data, but it is secondary to being found at all. The AI visibility area has a modest weight, and it covers whether AI crawlers may read your site and whether your content can be quoted. For an overview of all seven areas, see our article on the seven areas behind your SEO score.
Reading the sample report
The product page shows a sample report for a site named example-shop.com, with an overall score of 71, 412 pages and 84 issues. The area scores in the sample are: crawl and index 82, on-page 58, links 77, speed 64, structured data 41, languages 90 and AI visibility 66. Three top fixes are listed with the points each would add: the canonical that points to a redirected page, 9 points; duplicate titles on 64 pages, 5 points; AI assistants blocked in robots.txt, 2 points.
This example shows how to use the report as a plan. The structured data area is the weakest, but it is worth only 10 percent of the total, so fixing it helps less than fixing the canonical problem in the heaviest area. The fix list is ranked by points gained, so you do not need to do the weighing yourself: start at the top, fix, re-check and watch the score grow. The product page describes the same loop: top fixes show how many points each one adds.
Why your score moves when you change little
Several things can change the score without a big change in your site. Understanding them keeps you calm.
- The share changes when pages are added or removed. A problem affecting ten pages is a large share of a hundred-page site and a small share of a thousand-page site.
- Fixing a widespread issue gives a large jump. That is expected, because the share of affected pages falls sharply.
- A new template can introduce an issue everywhere at once. A theme update that adds a missing tag to every page will lower the score quickly.
- External factors change. Broken external links appear when other websites change, even if you touched nothing.
For this reason, compare audits over time rather than reacting to a single number. Plans that include re-checks and comparison make this easy, and the Pro plan adds alerts when the score drops.
Using the score well
- Treat the score as a progress meter, not a grade. It tells you whether your site is moving in the right direction.
- Fix by points per effort. Take the high-point, low-effort items first.
- Look at the area scores. A good overall score can hide a weak area that matters for your business, for example languages on a multilingual site.
- Re-check after each batch of fixes. This confirms that the fix worked and that nothing new broke.
- Combine the score with real search data. Search Console shows what actually happens in results, and the audit explains why. Our article on how to read an SEO analysis report describes the habit.
If your issue is that pages are not appearing in search at all, start with our guide to why Google is not indexing your pages, then return to the audit to confirm the fixes. And if you want to go through a full technical review of a WordPress site, our technical SEO audit guide is a good companion.
Plans and what they unlock
A free account runs one check and shows the report locked, because a deep crawl and AI analysis take real server time. A paid plan unlocks the score, every issue, the affected pages and the fixes. According to the product page, the Starter plan is from 12 euros per month with 5 checks, the Pro plan adds alerts when the score drops and comparison of audits with a PDF, and the Agency plan offers white-label PDFs and a lead-generation form for audits of clients’ sites. Each check covers up to 10,000 pages. The crawler is polite: it makes about three requests per second and follows your robots.txt.
Turning the score into a work plan
A score is useful only when it leads to a decision. Imagine a small company that receives a report with an overall score in the sixties. A simple planning session can turn it into three batches of work.
- Batch one: the heavy items. Everything at the top of the fix list that adds the most points, especially in the crawl and index area. These are usually few in number and often need only one change in a template or a setting.
- Batch two: the widespread warnings. Duplicate titles, missing descriptions or heading problems that repeat across many pages. They look dull, but their share of pages is large, so they move the score steadily.
- Batch three: the details. Notices, individual pages and low-weight areas. Do these when the first two batches are finished, or when you are editing those pages anyway.
After each batch, run a re-check. If the score rises by roughly the number of points the list promised, the fixes worked. If it does not, something else changed, or the fix was incomplete, and the affected-pages list shows where to look. Write down the date, the score and what you changed, so that after a few months you can see the whole journey instead of a single snapshot. Teams that keep this log find it easy to explain progress to managers and clients, because every point has a reason behind it.
Frequently Asked Questions
How is the SEO score calculated?
Each issue has a severity, critical 10, warning 4 or notice 1, multiplied by the share of your indexable pages it affects, softened so that one forgotten page barely counts. Seven area scores are combined by weight.
What are the weights of the seven areas?
Crawl 25 percent, on-page 20 percent, links 15 percent, speed 15 percent, AI visibility 8 percent, structured data 10 percent and languages 7 percent.
Why can a warning matter more than a critical issue?
Because the severity is multiplied by the share of pages affected. A warning on many pages can add up to more than a critical issue on a few pages.
Does a perfect score mean top rankings?
No. The score measures technical and on-page health. Rankings also depend on content quality, competition and authority, so use the score as a guide to what you can control.
Will the audit slow down my website?
No. The crawler makes about three requests per second and follows your robots.txt rules.
How is this different from Site AI Audit?
Site AI Audit checks website health, such as SSL, security, email and speed. Site SEO AI Audit goes deep into search: every page, every link, the sitemap, languages and AI search.
The Bottom Line
The SEO score in Site SEO AI Audit multiplies the severity of each issue by the share of indexable pages it affects, softens the result so isolated defects barely count, and combines seven areas by published weights. That design makes the number meaningful: widespread problems and problems in important areas move it most, and the ranked fix list shows how many points each correction adds. Use the score to decide what to fix first, re-check after each batch, and judge progress by trends rather than single readings. If you would like a specialist to turn the report into action, see our SEO services or ask for help directly from your report.