The working file on invalid traffic.

GA4 removes known bots automatically and won’t tell you how many. Everything else lands in your reports. How to spot it, what you can and can’t filter, and how to find the paid traffic behind it.

Invalid traffic drains budgets and corrupts the signals that decide where the next dollar goes. What it is, the forms it takes, how it is measured, and what a defensible count changes.

A share of every paid campaign’s traffic was never human, and the platform report will not tell you which share. This chapter covers the full spectrum, from crude scripted clients to sophisticated invalid traffic, and the architecture that separates bots from customers.

A share of every paid budget goes to visits that could never convert. What that tax is, why native, push and pop carry the most of it, and the one lever that cuts it.

Push networks often say their traffic is all human because every click comes from a subscriber. The subscription is exactly where the problem starts. Where invalid visits enter a push campaign, how to tell them apart from people who simply weren’t interested, and what to cut.

Most click-fraud tools sell detection and blocking as one product. They are two jobs, and on native, push and pop networks the second one mostly cannot be automated anyway. What blocking does well, where it misses and misfires, and how to cut bad sources on evidence instead.

Your invalid traffic (IVT) rate is the share of visits judged invalid. How to calculate it per publisher and placement, a worked example, and why your own baseline beats any published benchmark.

Google credits some invalid clicks on its own and reviews advertiser claims for the rest. This is the full sequence — verify the credit, build the evidence file, file the report, and cut the source with exclusions.

Where to exclude a bad zone, site, widget or publisher on every network we cover, what each lets you block, and the limits.

What 18 ad networks’ own terms say about money back for invalid traffic: who credits it, how to claim, and the deadlines.

Block low-quality Taboola sites: pass {site}, rank sites by invalid share, add them to Block Sites (1,500 per campaign), then re-measure.

Block bad MGID widgets: pass {widget_id}, rank widgets by invalid share, pause them in Widget optimization or import up to 1,000 UIDs per CSV.

Blacklist bad PropellerAds zones: pass {zone_id} and {subzone_id}, rank them by invalid share, and exclude them in Zone Limitation or by subzone.

Block bad ExoClick zones and sites: pass {zone_id} and {site_id}, rank them by invalid share, set Target Mode to Block and bulk-paste the zone IDs.

Exclude bad Zeropark sources and targets: pass {source} and {target}, rank by invalid share, Bulk actions → Pause, or block targets account-wide.

Run ValidVisit behind Voluum: where the script goes, the lander URL that passes zones through {var1}–{var10}, the click ID, and reading verdicts by zone.

Run ValidVisit behind RedTrack: where the script goes, the lander URL that passes MGID widget IDs through {sub1}–{sub20}, the click ID and reading verdicts.

Run ValidVisit behind Binom: script placement, the lander URL that passes ExoClick zone and site IDs through {t1}–{t10}, the click ID and Binom Protect.

Run ValidVisit behind BeMob: script placement, a lander URL that passes Zeropark source and target IDs, BeMob’s {clickId}, and its bot filters.

Run ValidVisit behind Keitaro: script placement for local and external landers, a lander URL for Taboola sites via {sub_id_1}, and {subid} as click ID.