As AI-fueled ad fraud scales to industrial levels, parts of the advertising stack are turning to blockchain for verifiable proof of who saw an ad and whether it led to a real outcome. In 2025, Google blocked or removed 8.3 billion ads, suspended 24.9 million advertiser accounts, and identified 602 million scam-related ads, underscoring how generative AI has made fake ads, fake users, fake clicks, and fake devices cheaper to produce and harder to detect. Alongside this escalating AI-versus-AI defense cycle, two blockchain-based models are emerging: Hakuhodo’s “Human-Verified Ad Network” that logs impressions to LG’s blockchain infrastructure after World ID verification, and Coinbase’s acquisition of Spindl, an on-chain attribution platform designed to prove that ads drive genuine on-chain actions.

Technology Overview

Google’s own defense posture signals the magnitude of the problem. Its Gemini system now analyzes hundreds of billions of signals in real time—account age, behavioral cues, campaign patterns—and stops more than 99% of policy-violating ads before serving. Yet the fraction that slips through still reaches users at scale, revealing the limits of probabilistic detection as adversaries deploy more advanced automation.

Against that backdrop, two complementary blockchain approaches have crystallized. The verified attention model—piloted by Hakuhodo in partnership with Tools for Humanity and LG Electronics—restricts delivery to human-verified users and writes an immutable receipt for each impression to LG’s blockchain. The verified conversion model—advanced by Coinbase through its January 2025 acquisition of Spindl—anchors ad attribution to on-chain events, creating a cryptographic chain of custody from a web click to a wallet interaction, an app install, a token purchase, or a staking event.

How It Works

Hakuhodo’s pilot in Japan (July–August 2025) integrated the agency’s “boba” mini-app with World ID and LG’s ledger. World ID is used to prove user uniqueness without revealing personal identity, ensuring ads are served only to human-verified participants. Each resulting ad impression is then recorded on-chain, producing a durable, inspection-ready record tied to proof-of-human status. According to reported outcomes from the pilot—spanning more than 3,500 participants and ten advertisers across electronics, travel, food, cosmetics, and education—the approach yielded a 50% increase in click-through rates and a 15-point improvement in bounce rates. For advertisers, the deliverable is a receipt confirming that a verified human received the ad, rather than a device or bot impersonating a person.

Spindl, founded by Antonio García Martínez—an early member of Facebook’s ads team—rebuilds attribution around verifiable events instead of cookies and probabilistic matching. Operating on Base, Coinbase’s Ethereum layer-2 network, Spindl traces the journey from the initial click to a measurable on-chain action, issuing a ledger entry that links spend to outcome. This architecture provides advertisers with attribution grounded in cryptographic verification rather than inference. Spindl maintains open standards for publishers and advertisers, seeking interoperability as on-chain commerce and wallet-based interactions become core performance signals.

Industry Impact

The scale of digital advertising raises the stakes for any verification breakthrough. Dentsu’s May 2026 forecast puts global ad spend at $1.06 trillion, with digital channels accounting for 69%. In the United States, IAB and PwC report that 2025 digital ad revenue reached $294.6 billion, with programmatic up 20.5% to $162.4 billion. The same automation that makes programmatic buying efficient also broadens the attack surface for fake inventory, simulated users, and fabricated outcomes.

Loss estimates illustrate the urgency. Juniper Research projects that global ad fraud losses will rise from $84.2 billion in 2023 to $172.3 billion by 2028 as adversaries learn to mimic human behavior and slip past detection. In connected TV environments, DoubleVerify found that bot fraud accounted for 65% of all fraud in 2024, with compromised devices imitating living-room viewers to trick measurement systems. When a purported household viewer is actually a botnet-controlled device, delivery metrics become contested claims rather than facts.

Blockchain’s offer to advertisers in this setting is a receipt: an immutable record of what the system observed, fixed at the moment of delivery or action. In the verified attention model, the receipt shows that a unique human saw the ad. In the verified conversion model, the receipt shows that the ad generated a concrete outcome. Together, these records form an auditable trail that can complement existing ad-tech systems where fraud risk and budget size justify more rigorous proof.

What Blockchain Cannot Do on Its Own

A blockchain reliably preserves inputs; it does not validate reality. The trust boundary is the oracle layer: confirming that the viewer was human before logging an impression, that the device was legitimate, that the impression was viewable, and that any downstream action was genuine. If the verification step is compromised, a fraudulent identity receives the same permanent record as an authentic one. World ID’s design attempts to address part of this by separating proof of personhood from personal identity, allowing uniqueness to be proven without revealing who the user is.

Adopting such systems invites regulatory and consumer scrutiny, especially in markets where biometric data collection is contested. There is also an ecosystem constraint: Google, Meta, Amazon, and major connected TV platforms control their own measurement stacks and have limited incentive to embrace a neutral, blockchain-based receipt layer that might dilute their hold on attribution. As a result, the most practical near-term path runs through segments where platform owners benefit directly from increased advertiser trust—crypto apps, independent CTV inventory, rewards campaigns, wallet-based commerce, and gaming.

Future Implications

Two trajectories are plausible. In a bull case, advertisers running high-value, fraud-exposed performance campaigns insist on verifiable logs that probabilistic methods cannot supply. Blockchain integrates with existing stacks as a parallel audit trail, and even a small redirection of spend away from fraud—1% to 3% of the $172.3 billion projected for 2028—would protect roughly $1.7 billion to $5.2 billion in value. In a base case, blockchain verification remains a specialized control for high-risk channels—commercially meaningful but not a wholesale replacement for measurement from entrenched platforms. In a bear case, incumbents improve AI-driven fraud detection quickly enough to keep verification and attribution fully in-house, narrowing blockchain’s role to crypto-native apps and limited proof-of-human pilots.

Either way, the direction of travel is clear. AI has made fake behavior inexpensive to generate and difficult to distinguish from legitimate activity, forcing detection systems into constant escalation. The Hakuhodo model demonstrates that on-chain, proof-of-human impression logs can work in live campaigns with mainstream brands. The Coinbase–Spindl approach shows that advertisers can link spend to on-chain outcomes with a verifiable chain of custody. A future conditional payout layer—settling advertising spend only when a verified event occurs via smart contracts or rules-based triggers—extends the same logic to settlement. Wallet-based targeting, which segments audiences by on-chain behavior rather than cookies or device IDs, adds another dimension to relevance without relying exclusively on legacy identifiers.

The likely next step is quiet integration rather than visible disruption. The infrastructure delivering these proofs does not need to be user-facing, and the end consumer may never know the underlying system is a blockchain. If advertisers ultimately decide that probabilistic measurement cannot be trusted in high-risk environments, on-chain proof systems for attention and conversion will be positioned to absorb the spend that seeks verification, even as platform-native AI defenses continue to evolve.