Is anybody even coming? We built a tiny traffic counter on our own site, switched it on, and here is the first honest read.

The quiet fear behind most 'is my site working?' questions is simpler than any dashboard: is anybody actually coming? I felt it too. So on our own site we put in a very small traffic counter — the kind an owner can understand without any technical background. This is what it does, what it deliberately does not do, and the first number it gave back — including why I do not treat that number as a headcount of people. No cookies, no third-party script, and nothing that hands your visitors off to someone else.

The usual way, and why I did not like it

The normal answer is a snippet you paste into every page. That snippet phones a company you do not control, drops a cookie, and needs a banner asking everyone's permission before it will count a single visit. It also mostly counts everyone the same way — a person, a search robot and an AI reader all look like one identical hit. It works, but it is a lot of machinery whose real product is other people's browsing. I wanted the smallest honest version of the same fact, on our own server.

What ours actually is

Ours is one small piece of logic on the server that reads the request the server was going to answer anyway. It adds no cookie, runs no extra script inside the page, and stores no personal address. The closest thing to an identity it keeps is a scrambled number that changes every day and cannot be turned back into who or where anyone is. For each visit it notes only the page, whether it looked like a person or a machine, and whether the page was served — then it writes that one line and forgets the rest.

Because it reads the caller's own badge — every program that asks for a web page announces what it is — it can separate three kinds of visitor: a normal person, a search crawler, and one of the AI readers that people keep telling you to let in. That separation is the useful part. A spike that is entirely machine reads is a completely different fact from a spike of people, and the honest tool is the one that tells you which you are looking at instead of just handing you a bigger number to feel good about.

The read side: counts only, and it hides from search

There is a second half, the part that shows the totals. Ours returns numbers and nothing personal: no address, no full badge, no visitor identifier. It will not answer at all unless you present a private key, and every answer it gives carries a quiet leave-this-out-of-search tag so the summary never turns up where a stranger might read it. The design point is narrow on purpose — the person who owns the site can see their own totals, and nobody else can pull that reader off the shelf.

The first read-back, and the reason it is not a headcount

Here is the part most write-ups would paper over, so I will give the reading and the weakness in the same breath. The layer is switched on now: the summary address, asked without the private key, refuses rather than answering — and asked properly, its first day of running recorded 45 requests classed as human across four per-day visitor keys, 2 classed as AI readers, 1 as a bot, and no referrers at all. Then the honest part. My own anonymous checks of this site present an ordinary browser badge, so they land in the human total. That 45 is an upper bound on people, not a count of people, and the tell is the empty referrer column: not one of those requests came from a search page or somebody else's site. I keep the number anyway, because a bound I can re-read is worth more than a feeling, and because the rule runs the same direction here as on a paying client's report — where I could not measure, the row says not checked rather than guessing.

What I can state, because it is measured and blunt, is the business-level fact: as of 2026-10-16 this service has received no enquiries and no paid orders at all. The site, the check and the payment path exist; the audience does not. That is the honest version of 'does anybody come to my site', and it is a far more useful place to start than a confident guess dressed up as insight.

What an owner can take from this

• Before you pick any counting tool, be clear what a view is. If it cannot tell a person apart from a machine, it will over-count the moment a crawler shows up, and you will make decisions on noise.
• Prefer the smallest thing that reads what your own server already sees over a snippet that ships your visitors somewhere else and then needs a permission banner to work.
• Whatever you use, open its summary and ask what it cannot tell you. Our own counter could not, today, hand me a real total — and admitting that is worth more than a confident dashboard.
• Split two questions that get muddled: are my pages reachable at all (a robots-and-headers check, about a minute of work), and is anyone actually arriving (a count, which needs a tool you trust). A traffic number answers neither of them until the plumbing underneath is sound.

Where the method comes from

Everything here is about our own site and our own code; the reads were anonymous and pointed only at addresses we control. The same care — the small file first, ordinary read requests only, a short pause between hits on one host, nothing executed, and never inventing a figure — is the method behind the one-time written report we run on up to twenty-five agreed pages of a single live site, priced at USD 99. If your real question is the reachability half — can anyone, human or machine, even get to your pages — that is the half this report lists line by line, with the observed value beside every finding so you can re-check each one yourself. You can see it at https://svc.openyuri.com/ .