Fraud & invalid traffic
Click farm
A click farm is a group of low-paid workers, or racks of real phones, used to click ads and perform searches by hand so the activity looks human.
Where bots imitate people, a click farm uses actual people or actual devices. Workers sit with rows of phones and follow a script: open the page, tap a related search, tap an ad, wait, repeat. Because a human hand and a real handset are involved, many automated checks pass.
In search arbitrage a click farm may be hired by a fraudulent publisher to click its own feed ads, or it may sit behind a "traffic seller" offering suspiciously cheap visitors. Google's definition of invalid clicks covers this directly: manual clicks intended to inflate publisher profits are invalid.
What gives farms away is sameness. Many devices share a few networks or locations, the same device models recur, sessions follow the same path and timing, device settings do not match the claimed country, and nobody ever buys anything from the advertiser.
Defence is about outcome and pattern, not single clicks: Post-click analysis, per-source reporting and refusing sources that cannot explain where their visitors come from. See incentivised traffic for the softer cousin, where real users are rewarded for clicking.
Think of it like this
It is a queue of paid extras outside a restaurant to make it look popular. They are real people, but none of them will order a meal.
Related terms
Click fraud
Click fraud is deliberately generating ad clicks with no real interest behind them, to earn publisher revenue or to drain an advertiser's budget.
Incentivised traffic
Incentivised traffic is visitors who came because they were rewarded with points, cash or perks, not because they cared about the page or its ads.
Self-clicking
Self-clicking is a publisher, or people acting for it, clicking the ads on its own pages to increase its earnings.
SIVT (sophisticated invalid traffic)
SIVT is invalid traffic designed to look like real people, which can only be found with advanced analysis that compares many signals at once.
Post-click analysis
Post-click analysis studies what visitors did after clicking, in aggregate, to find invalid traffic that looked normal at the moment of the click.