Which types of fraud are detected?

Ad fraud takes many forms, and no single check catches all of them. The Ad Fraud Protection page lists the types the traffic quality layer looks for.

Types detected

• Bot traffic: automated, non-human visits.
• Invalid traffic in general.
• Data center traffic: requests from servers rather than real devices.
• Proxy and VPN abuse, used to fake location.
• Click flooding: masses of clicks to steal attribution.
• Impression fraud: ads counted but never really seen.
• Attribution fraud: taking credit for conversions that would have happened anyway.
• Abnormal behaviour: patterns that don't look like real people.

It also stops fake domains, unsafe content and malware on advertiser landing pages.

Where each is caught

1. Before the bid: blocklists, junk and fake domain names, invalid and unmeasurable traffic. Requests from bad sources are refused before any money moves.
2. At the click: each click is checked and placements get a click quality score.
3. After the fact: sources are scored and the worst are blocked automatically.

Pre-bid decisions take milliseconds, so the checks don't slow the auction. The platform reports 9.5 billion auctions scanned monthly and more than 410 million malicious IPs blocked.

What advertisers can add

Rules can be customized per advertiser, campaign or traffic source, with whitelists for trusted partners.

Related: How AI-powered fraud detection works · Pre-bid filters: what is refused before the bid