Sites say a great deal about themselves to anyone who opens them, and the scan reads what is on offer. Every detection on the result page carries the evidence behind it, so each line can be checked rather than taken on trust. That is the part that matters, because a list of logos with nothing under it is a guess in a nice font.

Because it follows from something else rather than from evidence of its own. A site running Next.js is running React whether or not React names itself, and a site running WooCommerce is running WordPress by definition. Those are drawn quieter and labelled, because an inference and an observation should not look alike.

Three common reasons. Some technologies leave nothing visible from outside at all, which is why a site built on Go or Rust will only ever show the web server in front of it. Some appear only well after a page has started running. And some are simply not on the page you scanned: an analytics script that loads on the checkout is not on the front page. Nothing detected is not the same thing as nothing there.

No, and nothing can. When a site is proxied, every request goes to the proxy and the origin never appears, which is what the proxy is for. The page says the address belongs to Cloudflare rather than claiming the site is hosted there, because the front door is not the building. Occasionally the origin still names itself through the proxy, and where that happens you will see both: the CDN in front and the server behind it.

The network that actually answered, rather than a registry record or a bought data set. That is why the name reads like 'AMAZON-02' or 'HETZNER-AS' instead of like a brand: it is the network's own registration, and the country beside it is where that network is registered, which is not always where the machine sits.

It appears in their server logs like any other visit, named honestly: the scan identifies itself as Onymu's and links back to this page. It is lighter than opening the site in a browser. A site that would rather not be read can refuse the scanner in robots.txt or block it outright, and some do.

A scan is live. The answer is then held briefly, so the same site checked twice within a few minutes can repeat itself rather than being fetched again. A stack only changes when somebody deploys, so that is well inside the interval that matters for almost every site. The exception is your own, in the minutes after you have changed something.

Lines in the site's own robots.txt that name a crawler belonging to an AI company and disallow it everything. Since 2023 most large publishers have added them, so it is a quick read on whether a site wants its content in training data or in an assistant's answers. robots.txt is honoured voluntarily rather than enforced, so it records an intention rather than a guarantee.

That is a large part of what it is for. A name sold as a going concern comes with whatever is running on it, and the scan shows what that is: whether there is a real site or a parking page, what it was built with, who hosts it, whether the certificate is being renewed, and what the previous owner will still be able to see if a third-party script stays in place after the transfer. Pair it with the Whois lookup for the registration and the DNS checker for the zone.