What people write about our website: four e-mails under the microscope (part 2)

In part 1 we wrote about the people who send businesses e-mails about how bad their website is. More have arrived since. So we decided to publish them.

Not for entertainment. We took every message seriously, measured its claims against our own website, and wrote the result underneath. Names, phone numbers and addresses are gone, because this is not about particular people but about the way it is done. We make one exception with the fourth message. That one is not a person but a paid service, and anyone can verify its claims on its own website.

The fourth case: a test that does not rate your website

The last message is different. It claims nothing specific about our site, it only offers to run an analysis showing how well language models understand our services. The link leads to a service called gptsignal.sk. We ran the test.

The result was 26 points out of 100, and right below it three packages at 39.99, 71.99 and 111.99 euro.

It does not measure what you think

The key thing you will not learn from the e-mail is written on their own Technology page. Their probes ask assistants through public interfaces, roughly 600 questions across five assistants. They never fetch your website at all.

That means the result says nothing about the state of your page. It says what language models happen to remember about your company, in other words brand awareness. The e-mail, however, presents it as an analysis that will show what is worth explaining more clearly on your website. Those are two entirely different things.

We tried it on two well known websites

To find out what the number means, we ran the same test on two websites that need no introduction.

So no, it is not a tool that hands everyone a bad number. Wikipedia scored 98. But Martinus, one of the best known Slovak brands with millions of customers, got 25, practically the same as a small IT service shop from Liptov. If the test cannot tell those two apart, it says nothing usable about a local company.

What you will not be told

• The methodology is not published. The score has four components, but nowhere is it explained how they are calculated.
• There is no Slovak company behind it. The terms and conditions name ULTEH LLC. Neither the About nor the Contact page carries a single identifying detail, no company number and no registered address, although the domain is Slovak.
• The prices contradict each other. The test offers packages from 39.99 to 111.99 euro, while their own site advertises a package at 49.99.
• And one more thing. Across four pages of this service we counted 72 long dashes. That is exactly the sign we described in part 1 as the most reliable trace of text nobody read after the AI.

We are not claiming this is a scam. It is sales. You just need to realise that the number does not measure your website, and that you would be paying to improve it to a company whose address you never get to learn.

What we actually measured on our own site

So that this is not only about their claims, we measured our own website. Four pages, six screen widths from 320 to 1920 points, plus the contents of the code.

The one thing none of them found

The same measurement turned up something that genuinely bothered us. The blog listing downloaded the thumbnails of every article at once, 4.4 megabytes in two hundred and fifty requests. On mobile data that was needlessly heavy.

We fixed it the same day. Thumbnails now load only as you scroll near them.

That is twenty five times less data. And this is the whole difference. You find a real weakness by measuring, not by a sentence someone mailed to hundreds of companies at once.

How to check these messages yourself

You do not need to be technical. Three things are enough, each under a minute.

1. Look at the page source. Press Ctrl+U in the browser and search for the word viewport. If it is there, the claim about not being ready for phones falls apart.
2. Look at the e-mail headers. In Gmail use Show original, in Outlook use Message properties. The name of a mailing service in the return path, or hidden recipients, means the message was not written for you.
3. Open the sender’s own website. On a small screen too. If someone sells websites, their own website is the work sample. A free subdomain instead of an own domain says everything.

If you want hard numbers, put your address into Google PageSpeed Insights. It is free and within a minute you know where you stand. What a quality website should meet is covered in what a quality website should look like in 2026.

Conclusion

Four messages, four different senders, one pattern. The specific technical claims fell apart once checked, the rest were impressions, an offer to send specifics only after you reply, or a number from a test that never opens your website. Not one of them found the only thing on our site that was actually worth fixing.

This does not mean e-mail offers should never be trusted. It means you can verify a claim about your website before you pay for it. And when someone tells you something you can disprove in half a minute, you already know how much work went into the rest.

This article is part of our Business and IT overview.