Every search ad click costs an advertiser money. When the click comes from a bot, a tricked visitor or a hijacked browser, somebody in the chain keeps that money for nothing. The Fraud Desk maps every way it happens in search arbitrage, how each one is spotted, and what stops it.
A feed pays whenever a sponsored result is clicked. An honest publisher earns that by bringing interested people. A dishonest one looks for the cheapest thing that resembles a click, which is click fraud.
The click on the ad was never a real person choosing to click.
Software, not a person, clicks the sponsored listings on a search feed page so that the page owner earns a share of each paid click.
Rooms of low-paid workers, or racks of phones, click ads by hand so the clicks look human because they are.
The publisher, or friends and staff acting for the publisher, click the ads on their own pages to raise their earnings.
Visitors are rewarded with points, cash, prizes or unlocked content for clicking ads, so they click for the reward, not out of interest.
The page is designed so that people click ads by mistake, or cannot get what they came for without clicking one.
Clicks are fired in the background by an app, script or hidden page, without the user seeing or choosing anything.
The “person” on the page is software dressed up as one.
Bots route their visits through real households' internet connections so that each fake visitor appears to be a person at home.
Real browser software is driven by a script with no screen and no person, producing visits that pass basic "is this a browser?" checks.
Visits come from rented servers in data centres, where programs run but people do not browse.
A visit lies about what it is: it claims to be a new iPhone in New York when it is an emulator or cheap device somewhere else.
Real people, tricked into arriving or clicking.
Reviewers and real users are shown different pages, so the ad platform or search feed approves something that visitors never see.
The ad that buys the visitor promises something the landing page and the search ads behind it cannot deliver.
An ad or page pretends to be a tool, button or system message, so people click expecting something to happen that never does.
A page imitates a news site or invents a celebrity endorsement to give a commercial pitch borrowed credibility.
Ads, domains or keywords use a brand name to suggest an official connection that does not exist.
Bad traffic passed off as coming from an approved source.
Low-quality or fake visits are mixed into a stream of good ones so the average still looks acceptable.
The traffic lies about where it came from, naming an approved site as its origin when it really started somewhere else.
An approved feed partner quietly passes its feed to other parties, or brings in traffic from sources the search engine never approved.
Searches and visits forced by software the user did not choose.
Software changes a person's browser so their searches go through a monetised search page they never chose, or injects extra ads into the pages they visit.
A malicious ad or script sends the visitor to another page without a click, sometimes to a feed page, sometimes to a scam.
Visitors to a parked or expired domain are instantly forwarded to whoever pays most, including scams, with no page shown and no choice made.
Someone registers misspellings of popular web addresses and earns from the people who mistype.
Bending the feed’s rules to squeeze out more paid clicks.
Searches are sent to the feed that no person actually typed or chose, so ads are shown against invented intent.
The publisher loads the page with high-paying search terms that have little to do with what the visitor came for.
Feed units are loaded where nobody can see them, or piled on top of each other, so requests and clicks are counted without any real viewer.
Thin pages exist only to push bought visitors on to ads, giving the visitor nothing of value in between.
What it does to the advertiser who paid for the click.
Bots or paid workers fill in forms and trigger "sales" signals after the click, so bad traffic looks like good traffic.
An advertiser's search budget is spent on partner sites they cannot see or would not choose, where clicks convert poorly.
Every click on a search ad starts as a visit from some device. Most are people; some are scripts, crawlers or hijacked phones. IVT detection is the work of telling them apart.
TrafficGuard has identified that up to 34% of traffic from paid ads on ChatGPT may be invalid. This highlights significant concerns about traffic quality and potential ad fraud on AI-driven platforms.
Reports indicate that ChatGPT ads are now being targeted by invalid traffic and click fraud schemes. This development highlights the spread of fraudulent activity into new ad formats and platforms.
Reports indicate a notable rise in ad click fraud, affecting advertisers and platforms alike. This surge can undermine campaign effectiveness and lead to increased scrutiny of traffic quality across the digital advertising ecosystem.
Documented operations, what they did and what they taught the industry.
Zero-click parking abuse: visitors to parked and typo domains sold on to scams and malware
Infoblox (reported by KrebsOnSecurity)Brand-safety and transparency research on where search-partner ads appear (not a fraud ring)
AdalyticsCloaking software sold to advertisers to hide the real landing page from ad review
Facebook (now Meta)Browser-modifying malware that injected extra ads into search results
Microsoft 365 Defender Research TeamAd fraud run through malware-infected home computers and hijacked IP addresses
Google and White Ops (now HUMAN), with the FBI and US Department of Homeland SecurityAdvertiser class action over undisclosed ad placement on parked domains and error pages
AdWords advertisers (class action, Northern District of California)No. Search arbitrage sends real people to real ads under a contract, and the advertiser pays only if a person chooses to click. Ad fraud fakes the people or the clicks. The two overlap when an arbitrage operator, or a traffic seller it buys from, uses bots, click farms or deceptive tricks to inflate clicks. More
The main types are fake clicks from bots or click farms, paying or tricking people to click, laundering bad traffic through clean-looking sites, cloaking that shows reviewers a different advert or page, hijacked browsers and adware that redirect searches, forced or fake search queries, and fake conversions sent to make poor traffic look valuable. More
IVT stands for invalid traffic. It means ad clicks, impressions or visits that do not come from a real person with genuine interest. The term covers obvious automation such as crawlers, deliberate fraud such as bots and click farms, and human activity that should not be billed, such as accidental or incentivised clicks. Advertisers are not meant to pay for it. More
Invalid traffic is any visit or ad click that should not be charged to an advertiser: bots, repeated or automated clicks, people paid or tricked into clicking, and clicks made by accident. In search arbitrage it matters twice over, because the publisher has already paid for the visitor and will not be paid for the click. More
GIVT, general invalid traffic, is non-human activity that is easy to identify with lists and simple rules: known crawlers, data-centre addresses, declared bots. SIVT, sophisticated invalid traffic, is built to look like a real person: hijacked devices, residential-proxy bots, manipulated or incentivised human clicks. SIVT needs behavioural analysis and multiple signals to catch. More
Click fraud is deliberately generating clicks on pay-per-click ads with no real interest behind them, in order to earn money from the clicks or to waste a competitor's budget. It is done with bots, networks of infected devices, paid human clickers or a publisher clicking its own ads. Advertisers pay for it unless it is detected and refunded. More
In layers. Simple filters remove known bots and data-centre addresses. Device and browser checks look for automation and spoofing. Behaviour analysis examines timing, movement and click patterns. Statistical checks compare each source with normal rates. Finally, advertiser conversion data shows which sources produce clicks that never become customers. Checks run before the click, after it, and at month-end. More
A traffic quality score is a rating of how genuine and valuable a source of visitors is. Search engines keep internal scores that influence how much they pay for partner clicks. Independent vendors publish their own, built from signals about bots, fake clicks and behaviour. A low score usually means lower revenue per click, deductions or loss of the feed. More
ClearTrust, the company behind this desk, scores traffic quality with the TQI Score™: more than 150 filters for bots, fake clicks and fake impressions, rolled into one number a team can act on before a feed provider does.