AdSense invalid traffic — what triggers it and how to respond

Intent is not part of the test

The phrase sounds like an accusation, and publishers respond as though they have
been accused of fraud. Often they have not. Invalid traffic simply means clicks
and impressions Google will not pay for, and a well-meaning publisher can produce
them through an ad placement or a traffic source they trusted.

That distinction matters for the response. If you assume the finding is an
allegation, you argue. If you understand it as a measurement, you go and look at
what produced it.

The four things that produce it

Publisher clicks. Yours, your family's, your colleagues'. Including the ones
made to check whether the ad works. Google correlates by device, network and
behaviour, and a small site is the easiest case in the world to identify: it has
a handful of visitors and one of them keeps clicking.

Bought traffic. Traffic exchanges, paid-to-click networks, "10,000 visitors
for $5" packages, and social boosting from sellers who cannot say where the
traffic comes from. These deliver exactly what invalid traffic detection is built
to find, and buying them is the single fastest route to this notice.

Bots. Scrapers, uptime monitors, headless browsers, and unsolicited crawlers.
This category often arrives without the publisher doing anything, which is why
looking at traffic sources matters even when you have bought nothing.

Accidental clicks. The one that catches honest publishers. An ad unit
directly beneath a navigation item, adjacent to a download button, or dropped
into the reading flow on a phone produces mistaken clicks at scale. Every visitor
was real; every click was a mistake. The data cannot tell the difference.

Placement is a decision you can audit

Of the four, accidental clicks are the one entirely within your control and the
one most often overlooked, because a placement that seems fine on a desktop
behaves differently under a thumb.

Check where units sit relative to things people tap:

• Immediately below a menu or a breadcrumb
• Adjacent to pagination, a download link or a play button
• Between paragraphs at a point where a reader is scrolling quickly
• Anywhere a sticky element can shift underneath a finger mid-tap

A unit that earns well because it is easy to hit by accident is not earning; the
revenue is reversed later, and the pattern is recorded against the account.

Deductions are not a ban

Seeing earnings adjusted downwards is alarming, and at small scale it is routine.
Advertisers are not charged for invalid clicks, so the revenue is reversed.
Detection is retrospective, so a deduction can appear weeks after the activity
that caused it.

A deduction is a correction. A policy notice, or ad serving being restricted, is
an escalation. They are different events and it is worth not confusing the first
for the second.

If you receive a notice, answer with data

The response that works is factual. Say what you found, attach evidence, and
describe what changed:

• Analytics exports showing the traffic pattern and its sources
• The dates the anomaly began and ended
• The ad placements before and after, if you moved them
• What you have stopped doing, named specifically

A denial with no evidence is the weakest available reply. An account that
identifies its own cause and shows the fix is a much better position — even when
the cause turns out to be something you did.