Paying for bots
You think you are paying for bots. Now prove it.
Suspecting it is easy. Getting a network to act needs specifics: which clicks, at what times, from where, and why each one was judged automated.
Bot traffic is not one thing. Some of it is harmless automation that you still should not pay for — crawlers, uptime monitors, chat apps fetching a link preview. Some of it is deliberate: data-centre addresses dressed as phones, headless browsers driving through your funnel, one address clicking forty times an hour.
The reason it persists is that a total is easy to dismiss. "Thirty per cent of this looks like bots" invites a shrug. A list of two hundred clicks, each with its timestamp, address, network and the reason it was flagged, invites a credit.
The second reason is that most trackers hand you a score with no working. If you cannot say which signal fired, you cannot answer the obvious reply, which is that your filter is too aggressive.
How it works
The mechanism, not the marketing.
- 01
Address
Whether the click came from hosting rather than a home or a phone network, from a proxy, a VPN or a Tor exit.
- 02
Client
Whether the browser is being driven by software, has no user agent, or claims to be a crawler it is not.
- 03
Behaviour
Bursts from one address, clicks arriving faster than a person could make them, sessions with no lander click ever.
- 04
Verdict
The signals are weighed against a threshold you set, and the reasons are kept on the click.
What to check tonight
You can do all of this without us.
None of it needs an account anywhere. If it turns out your tracking is fine, you have lost twenty minutes and learned something.
Count clicks per address for one day
Sort yesterday by address. A long tail of ones and twos is people. A block of forties is not, and it is the first thing to screenshot.
Look at the lander click rate by zone
A placement sending hundreds of clicks where almost nobody ever clicks through your lander is either terrible traffic or automated traffic. Either way you should stop buying it.
Check the hours
Real traffic follows waking hours in the country it comes from. A flat line across the night from one source is a strong signal.
Split by connection type
If a source claims to be mobile but the addresses resolve to hosting providers, that mismatch is the single most quotable line in a dispute.
How UTMCAP does it
What you would get here.
Twenty-one signals, named
Each with a weight you can see, and two that cannot be switched off because they are not judgement calls.
The reason on every click
In words, exportable as CSV, which is the evidence a network argues with rather than around.
A repeat-address report
Built for exactly this: who clicked more than once, how often, and when.
Route them instead of paying
Send suspected bots to a harmless page rather than burning an advertiser’s offer on them.
Taken out of the money
Filtered clicks are excluded from cost and revenue, so your real CPA is not flattered or ruined by them.
The Copilot raises it
A campaign crossing thirty per cent filtered traffic over three days gets flagged with the figures attached.
Questions
The ones people ask next.
Read next
- What bot rate is normal?
- It depends entirely on the source. Search traffic is usually low single digits; pop and push can be far higher and still be worth buying. What matters is the rate for one source against its own history, and whether the good traffic still pays.
- Will a network refund me?
- Some will, especially with a click-level export and a clear pattern. Others will not, and the useful outcome is knowing to stop buying that placement.
- Do bot clicks use up my plan?
- Here, yes — the filtering is the work. They are excluded from your cost and revenue figures, so they do not distort what you decide on.
Point a domain at us and run a click through it.
The free plan takes a hundred thousand clicks a month and no card. If it does not do what you need, you have lost twenty minutes.