Who clicks your ads, and why
Contents
Share
Someone still gains from a click that never becomes a sale, and that someone is not always a competitor. The Yandex Direct antifraud guide and the Google Ads guide name four different parties. The report does not sign which one clicked.
Below is who benefits, and which trace that benefit can leave. The click costs you money only if it stayed in spend. That boundary is the subject of what ad click fraud is. Signs that do not try to guess a name are collected separately.

Software clicks with no person
Yandex classes clicks from bots and fraudulent software as fraud. Google describes the same thing as non-human traffic: programs that imitate an ordinary walk through a site. The gain does not belong to the program. It belongs to whoever started it. Someone else’s daily budget runs out early, or a publisher is paid for the visit.
A click the filter recognizes goes into invalid traffic and is not charged. A click it misses looks like an ordinary spend row. That row is not labeled “this was a bot”. How a program’s session on the page differs from a person’s is covered in bot or human.
A person is paid to click
Yandex’s second case is clicks by motivated users who complete tasks for money. Google writes about publishers who pay people to click so the publisher earns more. The browser and the hand are real. A check that looks only for software can let this person through.
On the ad network the publisher is the one paid for the visit. The placements report is the first place that interest can even show up: a large share of clicks from one site and almost no target actions. That is a reason to open the visits, not proof that a task was ordered. The task itself is not in the account.
A competitor wants your budget empty
Google explicitly includes competitors who click manually to raise your spend. There is no interest in a purchase. If Direct recognizes the click, you are not charged. If it does not, the cost stays in spend, and less budget is left for people who searched for you on their own.
No name appears in the report. A repeat with no lead looks like a competitor, like software, and like an employee who opened your own ad. The guides do not list employees as a separate kind of fraud. Direct does not filter a real but poorly matched click. For the budget, your own click is just as empty when it brings no lead.
An accidental click has no customer
Yandex separates accidental traffic from fraud: a repeated click on one banner, with no intent to deceive. Google cites accidental clicks from poor placement, when an ad sits too close to a button. The person who clicked gains nothing. For you the click is empty until the system classes it as invalid and takes it off the bill.
An accident does not have to happen again tomorrow. A campaign ordered against your budget does. One day cannot separate them. You need a period, and a comparison with leads. How to read the traces without naming a culprit is in signs of click fraud. How to estimate the rubles already charged, rather than guess a percent, is in the budget-loss calculation.
The report keeps a trace, not a surname
| Who | Why they do it | What you can see without inventing a name |
|---|---|---|
| Software | Drain a budget or inflate a publisher’s fee | An ordinary click with no lead, if the filter did not remove it. Removed clicks are not in spend |
| A person paid to click | The publisher earns from the visit | A placement’s share of clicks with no target actions. That alone is not proof |
| A competitor | So your budget runs out sooner | Manual repeats with no purchase. No name |
| An accidental person | For no gain | A short spike the filter may already have removed |
Past that, the name cannot be reconstructed. You can stop paying for the next similar visit on the site, and separately estimate what was already charged.
The visit check does not produce a surname and does not write “this is a competitor”. It looks at behavior on the page, the browser environment, the network, how often requests arrive, and the ad-click parameters. A doubtful visit can be sent to a short computational task instead of being closed on one odd signal. A real customer on an unfamiliar phone should not fall off together with an empty click. A hard refusal is still possible when the visit warrants it. The opposite limit is just as plain. A single click from an ordinary phone, with no run of repeats, can be indistinguishable from a customer, and Binetro does not promise to catch it. Money Direct has already charged is not returned.