Somebody hands you an SEO audit and the first thing you look for is the number. Seventy-two. Eighty-eight. Whatever it is, you have already decided how worried to be before reading a single finding. The free search visibility audit I shipped this week scores a site out of 75, and getting that to a round 100 would have taken about ten minutes of arithmetic. That is exactly why it does not. A score out of 75 is annoying. It does not fit the mental shortcut that says a good result is somewhere in the eighties, and every reader has to stop and work out what happened. Stopping to work out what happened is the entire point, because the thing that produces a denominator of 75 instead of 100 is a decision about what was measured, and that decision matters more than the score sitting on top of it.

What the number is hiding

The audit scores three dimensions. On-page carries a weight of 30 and measures meta completeness and quality across the tracked estate. Technical carries 25 and covers crawlability, speed, structure and correctness of markup. Content authority carries 20, looking at depth, coverage of the topic space and entity clarity. Three dimensions, 75 points, and a target stated for after implementation so the score can be re-run to show movement. There is a fourth dimension. AI and answer visibility carries a weight of 25, and the free package does not score it, because doing so takes a standing prompt set run against each engine and that prompt set is not in the download. So the arithmetic had two possible endings. Divide the three remaining weights into 100 and hand over a familiar-looking number, or leave the total at 75 and make the gap visible in the score itself. Rescaling is what almost every scoring system does, and it is quietly dishonest. A rescaled three-dimension score of 84 and a genuine four-dimension score of 84 look identical on a page and describe different sites. One of them has been assessed on whether it gets quoted back by an answer engine. The other has not been looked at on that question at all, and the number gives the reader no way to tell which one they are holding.

Absent is not zero

The second decision was what the report does with the row. Deleting it was the first attempt, and it was wrong for a reason that only showed up once someone tried to hand the report to a client. Cutting the row put the gap in the rubric and in the README, which are documents the person running the audit reads. It put the gap nowhere in the document the client actually receives, which is the only place it does any work. So the row went back in, marked Not measured, with a line saying that an unmeasured dimension is not a failing one. Scoring it zero would have been worse than deleting it. Zero is a claim about the site: it says we looked, and there was nothing there. Absent is a claim about the audit: it says we did not look, and here is what you would need to. Those two statements send a reader in opposite directions, and a scorecard that cannot distinguish between them is producing a number that reads as measurement while carrying an assumption.

	What it says about the site	What the reader should do
Scored 0	Assessed, and performing badly	Fix it, and expect the score to move
Not measured	Not assessed	Find out, before deciding anything
Row deleted	Nothing, and the reader never learns a fourth exists	Nothing, because they do not know to

Where the gap bites

The reason the fourth dimension is worth this much argument is that it is the one currently moving. On a retainer for Lethwei World, a Myanmar combat sports media brand, schema markup and structured FAQ content produced an A+ AEO score, a passing Google Rich Result Test, and 6,170 impressions inside AI-generated search results over three months. None of those three outcomes would appear in a three-dimension score. The site could have been assessed on on-page, technical and content authority, come back respectable, and the work that actually produced the impressions would have been invisible to the number the client was shown. That is the practical cost of a denominator nobody checks. It is not that the score is wrong within its own terms. It is that the work it cannot see is increasingly the work worth doing, a shift I wrote about in more detail in the search result nobody clicks anymore. The same blind spot runs through the defect classes the technical dimension does catch. An absent sitemap and a sitemap that exists but is never referenced in robots.txt both score as deductions, and both are usually symptoms of nobody owning the sitemap rather than of a technical failure. The score records the deduction accurately. It has no way to record that the cause was a vacancy on an org chart.

Check your last audit

Go and find the last audit anybody handed you, whether it came from an agency, a tool subscription or a plugin. Look for the denominator first, and if the document does not state one, that is your finding. Then work out what the score covers, dimension by dimension, and write down what it never looked at. Most tool-generated audits do not assess answer-engine presence at all, which means their score is a three-dimension score presented as a complete one, and you have been comparing sites on it. If the report scores a category zero, check whether anyone actually ran that check or whether zero is what the field defaults to when the crawler finds nothing to parse. The audit package is free and the product guide has the link, along with what it deliberately does not do. Run it or do not. The habit is what transfers, and it costs nothing to start: before you act on a score, make somebody tell you what it was out of.