Developing
The column does not name Lunio, uses one unnamed campaign's charts and proposes no replacement estimate.
In 60 seconds
- The $63 billion figure matches Lunio's 2026 Global Invalid Traffic Report, which analysed over 2.7 billion paid clicks and found 8.51% invalid.
- Fou argues bots on CPM buys are paid when ads load, so click-based site rates miss impression fraud: "they don't click".
- He says IP-based detection fails because botnets use residential proxies and some humans use VPNs.
- His charts cover one unnamed campaign; the invalid share was about 60% on January 8 and 80-85% in late January and early February.
Impact Layer
- What changed
- A well-known fraud researcher has publicly challenged the method behind a widely quoted fraud-cost number.
- Why it matters
- Budget decisions and vendor claims use the $63 billion figure, so how it was built affects how much fraud advertisers believe they face.
- Who is affected
- Media buyers, verification vendors and finance teams reviewing fraud assumptions.
- Martech stack area
- Analytics, Advertising
- Stack impact
- Ask verification vendors whether their invalid rates come from clicks or impressions and how they detect residential proxy traffic.
- Business impact
- Moderate Opinion column by the founder of FouAnalytics, a verification tool. The year of the charts, advertiser, exchanges and impression totals are not given.
- Related technologies
- What to watch
- A response from Lunio and independent impression-level studies.
On October 16, 2026 ad fraud researcher Augustine Fou argued in a State of Streaming column that the widely cited $63 billion invalid traffic estimate understates impression fraud on CPM ads. PPC Land reported it on October 21. Fou sells FouAnalytics, a verification tool.
The estimate
The $63 billion figure matches Lunio’s 2026 Global Invalid Traffic Report, published January 20, 2026. The column does not name Lunio; PPC Land linked it through quoted wording and press coverage. Lunio analysed over 2.7 billion paid clicks from August 2024 to July 2025 and found 8.51% invalid, with TikTok at 24.2%.
Fou’s three arguments
- Bots on CPM buys are paid when ads load, so they have no reason to click. Click-based rates miss impression fraud. “They don’t click.”
- IP-based detection fails because botnets use residential proxies and some humans use VPNs.
- Multiplying click-based invalid rates by ad spend is incorrect. He offers no alternative figure.
The evidence
His charts cover January 8 to February 7 for one unnamed campaign; the invalid share was roughly 60% on January 8 and about 80-85% in late January and early February.
Not stated
The year, advertiser, exchanges and impression totals, how Lunio weighted its 8.51% average, and whether 80% is typical of the market (Fou does not claim it is).
Our read
The argument about clicks versus impressions is worth testing in your own logs. One campaign is not a market estimate.
Sources
- PPC Land: Fou says $63bn invalid traffic estimate undercounts fraud on CPM ads (opens in a new tab)
Source for the column's arguments, figures and gaps.