The clicks you pay for that no human ever made

Somewhere in your campaign reports are clicks that no person made. Scrapers checking prices, bots validating links, click farms doing what click farms do, and stranger things besides — all of them billed at the same rate as a genuine prospective customer. Ad platforms filter what they can and refund what they admit to, but the platform grading its own homework is exactly the arrangement you should want a second opinion on.

Behaviour is the tell

A bot can fake a browser, a residential IP, and a plausible user agent. What it rarely bothers to fake is behaviour. Watch the session behind the click and the difference is unmistakable: a human scrolls unevenly, hesitates, moves the pointer like a hand moves; a script lands, fires whatever it came to fire, and leaves in a straight line. Sessions with no scrolling, impossible timing, or a viewport no phone has ever shipped read nothing like shopping.

This is why session-level auditing catches what click-level filters miss. The filter sees a request and has milliseconds to judge it. A recording sees the whole performance, and the performance is where automation gets lazy.

What to do about it

The first step is simply knowing the number. When bot sessions are separated from human ones, your conversion rate, bounce rate, and cost per acquisition all change — usually for the better, because the humans were doing fine and the scripts were dragging the averages.

The second step is acting on the pattern. Bot traffic tends to cluster in specific placements, publishers, and hours. With recordings as evidence, excluding a placement or escalating to the platform stops being an argument about a hunch and becomes a claim with receipts. You will not get every wasted cent back, but you will stop buying the same fake clicks twice.