Site Analysis Tools, Measured By Their Own Yardstick

Method, and the number we deliberately threw away

On 15 August 2026 we fetched each homepage from a server in Helsinki and recorded four things: HTTP protocol version, HTML weight, certificate lifetime, and whether the page still emits FAQ structured data.

One number was deliberately dropped from the table: server response time. Our measuring host is in Finland and Iranian sites are far from it, so that figure mostly measures geographic distance rather than server quality. Our own site sits behind a content delivery network with an edge close to the measuring point, which means the comparison would cheat in our favour.

We spell this out because knowing which measurement is invalid is half of what analysis is. Any report that shows you response times for Iranian servers taken from a foreign tool, without saying this, has handed you a number with no meaning.

The ten sites

Read that last column properly. Three sites still emit FAQ structured data and one of the three is us. Google stopped displaying it on 7 May 2026, and we cover where it sits in the overview of what actually moves the needle in SEO, yet it is still on our own homepage. A table whose author is not the best row in it is more likely to be telling you the truth.

What in this table actually matters

Three findings earn their place. The rest of the columns are decoration.

One site is still on HTTP/1.1. kwrank.ir is the only entry without the second version of the protocol, which means the browser collects page resources one after another instead of in parallel, and on a page with dozens of files the difference is visible. No Persian analysis tool we have seen reports this at all.

The weight spread is eleven to one. From 69 KB to 785 KB for pages doing broadly the same job. Unlike response time, HTML weight does not care where you measure from; 785 KB is 785 KB in Tehran too. It is the only column in this table you can compare without caveats.

Two sites hold long certificates. kwrank.ir and cando.ac carry 364-day and 359-day certificates issued before the new cap. That is neither an error nor a risk, but it means their next renewal gets at most 200 days, and if renewal is manual their cycle doubles.

Why a tool score does not track a Google ranking

Any tool giving you a number out of a hundred has to combine dozens of criteria using weights it chose itself. Nobody told Google those weights and Google has told nobody its own.

The concrete example is in the table. A site with 785 KB of HTML and a 359-day certificate ranks eleventh for a query where a 69 KB site ranks fourth. If the score were predictive, that order would be reversed.

So what is a tool for? Finding what is broken, not predicting position. A page that is not indexed, a redirect that loops, a duplicated title, an image with no alt text: a tool finds these in seconds and a human takes an hour. What a tool cannot tell you is whether your page answers the question that was asked, and that is the thing that moves rankings.

A checklist you can run yourself

If you want a basic analysis with no tool at all, these four commands tell you more than most automated reports:

curl -sI https://example.com/ | grep -i "^HTTP\|content-encoding"
curl -sI http://example.com/ | head -n 1
curl -s https://example.com/ | wc -c
curl -s https://example.com/robots.txt | head -n 20

The first gives protocol version and compression, the second whether http redirects, the third the real HTML weight, and the fourth whether you have accidentally closed something off to Google yourself. The last one matters more to rankings than the other three combined and appears in no score anywhere.

What our own tools do and do not do

We run an SEO toolset too, so it is only fair to hold it to the same standard. Ours shows scores as well, but it does not invent one: what you see are Lighthouse's own category scores, which are Google's numbers rather than a formula we chose the weights for. The rest of the output is raw values: current position for a phrase, real search volume, certificate and DNS state.

Here is one of its limits, taken from the source itself. In the page analysis file, next to the metrics block, there is a note recording that INP is a field metric requiring real user data, so a lab test cannot measure it at all and TBT is reported instead. That sentence is written into the code so that whoever reads the output is not misled by it.

The bigger limit deserves saying too: none of them measures content quality and none of them can. No tool anywhere can. Anyone claiming their tool scores content quality is mistaking word count and keyword density for quality. The list of SEO tools and what each one measures shows which are free and which are paid.

Questions people ask

Is a free site analysis worth running? For finding technical faults, absolutely. For deciding what to write or why you are not ranking, no.

Why do two tools give the same site two different scores? Because their weights differ and both sets of weights are arbitrary. If you got two different scores, neither is wrong and neither means very much.

Does Google look at these scores? No. They are produced by the tool vendors and Google never sees any of them.

If what you want to know is where your site actually sits in the results and how far the first page is, that is not measured by a score but by a position, and checking your ranking for your own phrases gives you exactly that number. For fixing what an analysis turns up, technical and content SEO work is what we do, and if the site has no sound structure underneath, none of it matters as much as rebuilding the site with the structure right.

The measurement was taken on 15 August 2026 from a server in Helsinki and is repeatable with the four commands above. The ten sites come from Google's live results for that query in rank order and were not hand-picked.