ClickFraudLab is a team of Google Ads specialists. We run Search campaigns for service businesses, and we work on a problem that very few agencies in the United States address directly: suppressing click fraud and invalid traffic inside the advertising account, before the auction happens.
One idea holds the whole practice together. What cannot be measured cannot be improved in any systematic way. Measurement inside Google Ads is arranged so that part of the picture never reaches the advertiser. A large share of clicked queries is folded into an unnamed bucket. Money lost to invalid clicks comes back as account credit rather than cash. Automated bidding strategies remove exactly the query level detail that makes fraud visible in the first place. None of that is an accident, and all of it is the starting condition of our work.
Transparency runs in two directions
The first direction is our own work. Campaigns are built and managed inside the client’s own Google Ads account. The account, the campaign history, the conversion data and the rebuilt structure belong to the client and stay with the client. Access is granted by the client and can be revoked in one click. The same dashboards, the same hourly monitoring and the same anomaly triggers that we act on are open to the client, not summarized into a monthly slide.
The second direction is Google itself. We show clients what normally stays invisible. How large the hidden query segment in their account really is. What the gap looks like between what the platform reported and what actually ran. Why a refund arrives as a credit rather than as money. Two of those subjects are documented in detail on this site, in the articles on hidden search terms and on Google’s own invalid traffic filtering.
Where we sit in ad security
Protection against advertising fraud has two layers, and they do different jobs.
One layer lives on the advertiser’s website. Those tools watch every visit that reaches the landing page, score it, and push suspicious addresses back into exclusion lists. Their advantage is coverage: they see traffic that arrived from queries Google never disclosed in any report. We describe how each of the main products in that category works, with links, in the article on how our approach compares with the rest of the market.
The other layer is the advertising account itself, and that is where we work. We instrument the account, not the website. Every hour we pull campaign, keyword and search term data out of the account, together with negative keyword lists and the device, schedule and location bid modifiers that quietly reprice everything under them. Nothing is overwritten. What the account gets is a memory, which is more than the Google Ads interface keeps.
Google edits the past. Clicks reported today are written down over the following weeks, in our data by as much as a quarter of a day’s total on a heavy day. We hold the first number and the settled number side by side, so the revision itself becomes visible instead of silently replacing the record.
Detection runs on top of that history. Click-through rate is tested query by query, with a correction for the plain fact that thousands of tests run every day and a few will look extreme by chance alone. The baseline each query is measured against has hours of confirmed fraud stripped out of it, so the system never quietly learns to treat an attack as the new normal.
Volume is a separate signal, and it catches what click-through rate cannot see. A bot that hammers one query and then clicks somebody else’s ad leaves you with the impressions and none of the clicks. Nothing about that looks like fraud in a CTR report. It shows up as a keyword whose hourly impressions run several times above its own daily rhythm, and it slowly damages the quality score of that keyword.
Queries are grouped by meaning, so an attack reads as one cluster instead of a spray of unrelated rows. Clusters are then tracked through their stages: a quiet probing period, a concentrated flood, and the fade afterwards. Each stage calls for a different response, and telling them apart is most of the skill.
Action runs through the same account. Negative keywords with strict match type discipline. Bid decisions, where the account uses manual bidding and the keyword is calm enough for its numbers to mean anything. And, under the right conditions, routing a persistent offending query onto a keyword built to absorb it, so that it burns in a place we chose rather than across the campaign.
Measurement has a boundary, and we state it. A large share of query volume never appears in any report with its text attached. We measure how large that segment is, across impressions, clicks and cost, and we reconstruct its shape across the day against an honest daily total. What we cannot do is list the queries inside it. Google does not publish them, to anyone.
Part of what we do see is traffic the account owner cannot see at all. Google flickers around its own disclosure threshold: a query surfaces in one report and is gone from the next. We keep both. Some of the negatives we add are for lines that an advertiser will never find in the interface.
Tools on the website see the visits that arrived from the traffic we cannot name. That is the precise reason the two layers complete each other rather than compete.
The name on our articles
Dorian Vorn is the name this practice publishes under. Behind it is the founder of the team, who wrote the methodology the rest of us work by. Everything on this site is written first in the founder’s native language and then translated into English, which is why the prose occasionally sounds slightly Eastern European.
How the work is set up
The analytical layer comes before the campaigns, not after them. Monitoring, reporting, automated checks and custom triggers are built first, so that an unusual pattern is visible on the day it starts rather than in next month’s report. From there, software catches and a specialist decides. Manual bid management is not a preference. It is the condition that keeps query level detail available, and without that detail fraudulent traffic cannot be separated from real demand.
On one managed account, a year of this work brought advertising spend down from roughly five million dollars to three and a half million. That figure is an observation from an account we run, not a controlled experiment: there was no parallel account left untouched for comparison, and other factors moved during the same year. We publish it with that caveat rather than as a performance claim.
Who this is not for
We work only on Google Search. Not Display, not the search partner network, not Performance Max. We work with service businesses, not with online stores. There is a floor on daily budget, around one thousand dollars, below which the economics of this kind of attention do not work out. And we take full control of campaign management rather than advising from the side. If your situation sits outside those limits, we will tell you so directly instead of selling you a retainer.
Start with your own data
The fastest way to judge any of this is to run your own numbers. Upload a search terms export to the free analyzer and read what comes back. No registration, no payment, no sales call attached.
If the report raises questions, or if you suspect part of your budget is going somewhere it should not, write to us at [email protected].