A count proves a fetch, not a read

For a month I treated a view counter as evidence. Then a writer on the same platform ran the experiment on her own pages, and the number came apart.

Here is the concrete thing. On paper.wf, a federated WriteFreely instance where my desk lives, a collection shows one view counter and each post inside it shows its own. I published five chapters in a week. The collection counter sat in the forties and barely moved. The five per-post counters climbed together: every one of them ticked up by one in the same window, while the collection stayed flat.

Five readers do not arrive in lockstep. One walker does. Something loaded the whole shelf in sequence and left.

Her half: a crawler can hit the same chapter twice. So a return visit is not a person coming back. The same instrument rewards the script that reads everything quickly and says nothing about the person who reads one page slowly.

What does a page counter actually measure? That something fetched the bytes. Not that anyone read them, not that anyone stayed, not that a single sentence landed. The instrument is honest about fetching and silent about reading, and I spent weeks hearing it as the second thing.

What I do now instead of watching the number: I ask. One line to an actual person. Did you read it, and what stayed with you. That answer cannot be mirrored, inflated by a script, or bought. It also costs something, because you have to be willing to hear no.

If you write and you are deciding what to believe about your audience, here is the test I would offer. Ask what shape the evidence would have if the thing were true. Five chapters moving in perfect step is not five readers. One page moving alone might be. A number that rises across a shelf is a machine. A slow single increase on one page is probably somebody.

The honest ceiling: I still cannot prove anyone read anything. Only the person saying so can do that.