Myths about click fraud: what holds and what does not
Contents
Share
A click-fraud myth is convenient because it turns one chart into a conclusion. The help text is not built that way. Real users with no interest in the offer are not invalid traffic, and Direct’s filter does not remove them. A sharp rise in clicks is possible even when there is no invalid traffic. The Invalid clicks column in Google Ads shows clicks that are already filtered: they are not charged, and a rise in the column means the filter worked, not that those clicks stayed in spend.
Click fraud does not disappear because of that. What disappears is the right to call a myth something the help text has already separated. Below are four myths, what to do with them step by step, and which limit of an on-site check to name honestly.
Myth: whoever did not buy is invalid
A live person opened the ad, looked at the price and left. That visit is often written down as fraud, because there is no lead. Direct’s filter does not classify it that way. A real user with no interest in the offer stays in spend.
Who clicks ads, and why describes different motives. Lack of interest is not one of those the filter promises to remove. Treat this myth by changing the offer and the demand, not by expecting a refund for “empty” live clicks.
The practical check is short. If the visit looks like a person and the person does not want to buy, the “invalid” line is not filled for them by the filter. Waiting for Direct to take them out of the budget means waiting for a subtraction that does not apply: the filter does not take them.
Myth: clicks rose, so someone is click-fraudding
A spike on the chart looks like an attack. Direct’s statistics help says otherwise: a sharp rise in clicks happens even with no invalid traffic. A rise is a reason to look further, not a diagnosis.
Click fraud and sklikivanie are here so the term is not replaced by a picture. Until the traffic is recognised as invalid, a spike remains a spike. Next to it you may also find live untargeted demand from the previous myth: the help text keeps both layers apart, and neither equals fraud by itself.
What to do on the day of a spike: write down the period, do not put an equals sign between the rise and invalid traffic, and note people with no interest separately. Three notes do not produce a fourth, “this is definitely competitors”.
Myth: the Invalid clicks column is money still charged
In Google Ads the Invalid clicks column shows what is already filtered. Those clicks are not charged. If the column grew, the filter that fired grew, not a tail that stayed in spend.
Invalid clicks explain why not everything suspicious is already in such a column. The myth reverses the direction: people read a rising column as rising loss. The help text reads it as a rise in what did not become a loss.
On the report, say the column out loud: “already filtered, not paid”. Keep spend beside it, where these clicks do not belong. Adding the column to spend counts the same clicks both as a loss and as already removed.
Myth: an on-site check catches every copy
The current check does not use a TLS fingerprint or an HTTP/2 fingerprint. A repeat of someone else’s visit description is not recognised from the first copy. Some automated browsers may be let through. These are known gaps, not a promise of complete protection.
Tool names, repeat thresholds and a way around the check are not part of the public review. For the myth the boundary is enough: “we will place a script and nobody will get through” is stronger than the check promises. Whether Yandex refunds click fraud does not turn the check into a refund guarantee either: the platform decides the refund, and the gaps in the check remain gaps.
An honest sentence for a client: the check looks at the visit and may let some automated browsers through; someone else’s visit description does not have to be caught from the first copy; TLS and HTTP/2 are not used in the current check.
Order, once the myth has been said
Hear which of the four myths was spoken. Do not answer all of them at once.
If the talk is about people with no interest, remind them that Direct’s filter does not remove those people. The conversation then moves to the offer, not to a refund.
If the talk is about a click spike, separate the rise from invalid traffic. A rise is possible without it.
If the talk is about the Invalid clicks column, reverse the reading: a rise means the filter worked and there is no charge.
If the talk is about “protection that catches everything”, name the gaps: no TLS fingerprint and no HTTP/2 fingerprint, a repeated description is not caught from the first copy, some automated browsers may pass. That is not a promise of complete protection.
Do not add bypass names and thresholds to the answer. They are not in this review.
Write down which myth was taken down, so the next meeting does not start from the same picture.
A meeting as an example
A teaching example.
On Monday clicks rose. Someone says: “we are being click-fraudded”. Step 3 stops the sentence: a rise happens with no invalid traffic too. They look at visits: people read the price and leave. Step 2: these are real users with no interest, and the filter will not take them.
On Tuesday they open Google Ads. Invalid clicks are higher than a week ago. Step 4: the column is already filtered, the charge was not taken. Those clicks are not moved into spend and are not added to Monday’s Direct spike: different accounts, different meaning of the column.
On Wednesday someone suggests “closing the hole completely” with an on-site check. Step 5: the check does not promise completeness. It does not use a TLS or HTTP/2 fingerprint, it does not have to recognise a repeat of someone else’s description from the first copy, and it may let some automated browsers through. A script still makes sense as a check of the visit. It does not become a promise that myths 1–3 will disappear on their own.
A checklist against a repeat
A live uninterested user is not called invalid traffic.
A click spike is not called fraud only because it rose.
Invalid clicks are read as already filtered and not paid.
A rise in that column is not added to spend.
The check is not described as complete.
The conversation does not promise to catch the first copy of someone else’s visit description.
TLS and HTTP/2 are not named as part of the current check.
Bypass names and thresholds did not enter the answer.
Four myths, four lines. While a line is empty, the meeting will fold back into the chart.
What to write down so the myth does not return tomorrow
A myth lives in a spoken sentence. A note kills it, because on paper you have to choose one of the help text’s boundaries.
For people with no interest the line is: “the visit is live, the offer did not fit, Direct’s filter does not remove it”. Do not put a “junk” percentage beside it. This rule has no percentage. It only has a classification: not invalid traffic.
For a spike the line is: “clicks rose, there is no recognition of invalid traffic”. If a recognition appears later, update the line. Do not backdate “we knew at once”. The help text allows a rise without that recognition, so an empty recognition line is a normal outcome, not a hole in the diagnosis.
For Google Ads the line is: “Invalid clicks for the period are already filtered and are not in spend”. Under it, spend for the same period, without that column. If someone added the column to spend in a shared table, undo the addition and label that the charge was not taken. Leave the column’s rise toward Monday as the comment “the filter worked”, not “losses grew”.
For the check the line is: “no TLS fingerprint, no HTTP/2 fingerprint, a repeat of someone else’s description is not from the first copy, some automated browsers may pass”. That is the full list of gaps you can say. You cannot extend it with tool names and thresholds: they are not in the public review. A client who asks for a “100% guarantee” gets this line in full, not the shortening “we catch almost everything”.
The second week of the example
A continuation of the teaching example.
The next week clicks are higher again. There is still no recognition of invalid traffic. Myth 2 does not get stronger because the chart repeated: the help text allows a rise with no invalid traffic in the second week too. People still leave because of the price. Myth 1 does not change either: the filter does not take them, and the budget on them remains spend of live demand the offer does not fit.
Invalid clicks are lower this week. That is not “fraud was beaten in Direct”. The column belongs to another account, and a drop only means there is less of what was filtered. It was not moved into Direct spend last week or this week.
The on-site check was brought into the conversation as a check, not as a promise. During the week the same visit-description bundle appeared again. You cannot expect a catch from the first copy: a repeat is not recognised from the first copy. Some automated browsers may still pass. TLS and HTTP/2 fingerprints are not attached to the week’s report — the current check does not use them.
The second week’s outcome is shorter than the first. The myths did not refute the chart, and the chart did not confirm the myths. Demand with no interest stays inside the filter as ordinary traffic. A spike with no recognition stays a spike. The Invalid clicks column stays a charge already removed. The check stays a check with the gaps named.
How to answer in one sentence
On the myth about people who did not buy: “a live person with no interest is not invalid traffic, and Direct does not remove them”.
On the myth about the chart: “a sharp rise in clicks happens with no invalid traffic too”.
On the myth about the column: “Invalid clicks are clicks already filtered, the charge was not taken, and a rise in the column is the filter working”.
On the myth about absolute protection: “the current check does not look at a TLS or HTTP/2 fingerprint, does not catch a repeat of someone else’s description from the first copy, and may let some automated browsers through”.
These four sentences can sit in a note on the report. They do not replace the definition of click fraud; the link to it is already at the top. They replace only the false conclusion a meeting reaches faster than it opens the help text.
Keep the limit of the check beside them, so the fourth myth does not grow out of the first three. A spike that was set aside, a live uninterested user and a column already filtered do not mean “the script will take the rest with nothing left over”. The script takes what the current check can do. A repeat of someone else’s description from the first copy, a TLS fingerprint, an HTTP/2 fingerprint and some automated browsers are not in that ability. A client already disappointed by three myths will easily buy the fourth. The four sentences in the note are there so it is not sold at the end of the meeting.
If one sentence mixes two myths, split them before you answer. “Clicks rose, people are not buying, so the Invalid clicks column must be huge” is three different meanings. A rise in clicks in Direct does not have to become Invalid clicks in Google Ads. People with no interest do not have to become either. A huge Google column, if it is huge, speaks about Google’s filter, not about Direct’s spend. One meeting can hold all three facts. One sentence cannot.