Short answer: Every finding in Site SEO AI Audit has a severity: critical counts 10, warning 4 and notice 1. The score penalty is the severity multiplied by the share of indexable pages affected, raised to the power 0.6, so one forgotten page costs very little and a problem on 30 percent of pages costs a lot. The fix list is sorted by impact, the points you would gain by fixing that item alone, and then by severity. Work from the top of the list, and treat widespread notices as seriously as rare warnings.
A report with fifty findings can feel like a mountain. Which one first? The labels critical, warning and notice give a hint, but they are not the whole story. A critical issue on one page may matter less than a notice that appears on every page. The audit knows this, and it shows the answer as a number of points next to each fix. This article explains how the numbers come about, so that you can trust the order of the list and know when to override it with your own judgement.
The three severity levels
- Critical (10). The problem can stop pages from ranking or from being reached: pages with errors, a blocked website, a missing title, a canonical that points to a dead address, broken internal links, a very slow mobile score.
- Warning (4). The problem hurts visibility or trust, but pages still work: redirect chains, duplicate titles, thin content, missing descriptions, broken outbound links, hreflang errors.
- Notice (1). A minor weakness or a thing to confirm: a canonical to another page, missing alt text, a missing favicon, no llms.txt.
The weights are fixed, so two sites with the same problem get the same severity. What differs from site to site is how many pages the problem touches.
The share of pages matters
The penalty of a finding is its severity multiplied by the share of indexable pages that are affected, raised to the power 0.6. Problems that concern the whole website, such as a blocked robots.txt, count as affecting all pages. The exponent softens the curve in a useful way. The table shows what a few cases look like.
| Finding | Pages affected | Penalty |
|---|---|---|
| Critique | 1 of 100 (1%) | about 0.6 |
| Critique | 10 of 100 (10%) | about 2.5 |
| Critique | 30 of 100 (30%) | about 4.9 |
| Avertissement | 10 of 100 (10%) | about 1.0 |
| Notice | all pages (100%) | 1.0 |
| Critique | whole website | 10 |
Look at the fourth and fifth rows. A warning on one page in ten costs the same as a notice on every page. And a critical issue on a single page costs less than one point, which is why a lonely error page does not sink your score. The numbers show the idea of the formula: one forgotten page barely counts, and a problem that is widespread counts a lot.
From penalties to area scores
The penalties are added up inside each of the seven areas. The area score is 100 minus the sum of penalties, measured against 35 percent of the largest possible penalty for that area, and never below zero. Then the seven area scores are combined with their weights: Crawl and index 25 percent, On-page 20, Links 15, Speed and Core Web Vitals 15, Structured data 10, Languages 7 and AI visibility 8. If the Languages area does not apply to your site, it is left out, and the others share its weight, as the article on scoring one-language websites explains. One more rule: if robots.txt blocks the whole website, or no page is indexable, the total is at most 20. The complete description is in the article on how the audit score is calculated.
What “impact” means
Next to each finding in the top fixes, the report shows a number of points, such as plus 9 or plus 2. This is the impact: the number of points that the total score would gain if that finding alone were fixed. The audit calculates it by scoring the site again without that finding. The list is sorted by impact first, and by severity when impacts are equal.
This has two consequences. First, the order of the list is the most efficient order for raising the score. Second, severity does not always decide the position. A widespread notice can rise above a rare warning, because the number of pages it touches makes its penalty larger. The article on using your audit fix list shows how to work with it.
When to follow the list, and when not
The impact list is a good default, but it knows nothing about your goals. Use it with these questions.
- Is anything blocking the whole site? A blocked robots.txt or a site-wide noindex comes first, whatever the list says, because nothing else matters until it is fixed.
- Is the fix cheap? A template change that clears forty pages in ten minutes beats a manual edit of five pages, even if the impact is similar.
- Does it affect your money pages? A warning on the pages that bring customers can be worth more to you than a higher-impact item on pages that nobody visits.
- Is it intended? Some findings are choices, such as noindex on a thank-you page or a canonical to a filtered view. Confirm and move on.
- Is it risky to change? Large redirect changes or template rewrites need testing. Schedule them, instead of rushing.
A triage routine in four passes
- Pass one: emergencies. Look for site-wide critical findings and anything that explains an unusually low score.
- Pass two: the top five by impact. Fix them, preferably at the template or setting level.
- Pass three: the rest of the warnings. Group them by area and clear them in batches.
- Pass four: notices. Handle them in bulk, or accept the ones you chose on purpose.
After each pass, press Re-check and see how the score moved. Starter, Pro and Agency include re-checks. The article on the cycle of fix, re-check and compare explains how to make that loop short.
Common misreadings
- “All criticals are urgent.” A critical on one page of a hundred has a small effect. Fix it, but do not drop everything.
- “Notices can be ignored.” A notice on every page may add up to more than a warning on a few. Look at the number of pages and the impact.
- “A perfect score is the goal.” A score above 90 is excellent. The goal is a site that works for people, and some notices are choices.
- “The score is a ranking prediction.” It measures the technical and content health that supports ranking. It does not promise a position.
Use the numbers to explain priorities
The impact numbers are also a good way to explain priorities to a client, a manager or a developer. “This change adds nine points, and that one adds two” is clearer than “this one feels important”. Share the report as a link or, on Pro, as a PDF, and let the numbers carry the argument. The plans on the pricing page include enough checks to measure the effect of each batch of fixes.
A worked example
Imagine a site with 200 indexable pages and three findings. The first is a critical: 6 pages return an error. That is 3 percent of pages, so the penalty is ten times 0.03 to the power 0.6, which is about 1.2. The second is a warning: 40 pages have no meta description, which is 20 percent, so the penalty is four times 0.2 to the power 0.6, about 1.5. The third is a notice: every page lacks a canonical, which is 100 percent, so the penalty is exactly 1.0. The three penalties are close to each other, and the order of the list will depend on the areas they fall in and on the points each fix adds to the total.
What does the owner do? The error pages are cheap to fix, because six addresses need a decision. The descriptions are a bigger job, but a template can write a first version for all forty pages. The canonical is the cheapest of all, because one plugin setting clears the whole notice. A sensible order is the canonical setting, then the six errors, then the descriptions. The list by impact would give a similar answer, which is why it is a good default. And if the owner knows that those forty pages are tag archives that should not be indexed anyway, the description job disappears, because the better fix is to keep those pages out of the index.
Related reading
- How the Site SEO AI Audit Score Is Calculated
- Fix SEO Issues by Impact: Using Your Audit Fix List
- Fix, Re-check, Compare: A Repair Cycle for Your SEO Score
The bottom line
Severity says how serious a problem is, the share of pages says how widespread it is, and impact combines the two into points. Start with anything that blocks the whole site, then follow the list by impact, use your own judgement for choices and risks, and re-check after every batch.
FAQ
Why is a notice above a critical in my list?
Because the list is sorted by impact first. A notice that affects many pages can cost more points than a critical issue on a single page.
What does the exponent 0.6 do?
It softens the effect of the share of pages: one forgotten page costs very little, while a problem on 30 percent of pages costs a lot.
Does fixing everything give 100?
If no findings remain, the score approaches 100. Some notices may be intended choices, so a score in the nineties is already excellent.
Are the points in the list exact?
They are calculated by scoring the site again without that finding, so they are exact for the current report. After other fixes, the numbers for the remaining items can change.
Which plan shows the impact list?
Any paid plan shows the full report. A free account runs one check and shows the report locked.


