Daily Active Addresses vs. Unique Users: How Bots Distort Blockchain's Core Growth Metric
On February 25, a widely shared post on CoinSpectator flagged a growing structural problem: as AI agents gain the ability to autonomously operate crypto wallets, the already blurry line between human users and bot-generated activity on blockchains is about to get significantly harder to draw. For allocators evaluating blockchain adoption, this is not an abstract concern. It goes directly to whether the most commonly cited growth metric in digital assets, daily active addresses, actually measures what people think it measures.
The short answer is that it does not, at least not without significant adjustment.
What Daily Active Addresses Actually Measure
A daily active address (DAA) counts unique blockchain addresses that either sent or received a transaction on a given day. Glassnode reported Bitcoin at 667,518 daily active addresses as of February 25, 2026. On Ethereum and its layer-2 networks, combined active addresses regularly run into the millions.
The metric is appealing because it is easy to pull and provides a time series that looks comparable to traditional software's daily active users. But the comparison is misleading. In traditional software, one user typically maps to one account, enforced by email verification or single sign-on. On blockchains, there is no identity layer. One person can control hundreds of addresses at near-zero cost, and one address can represent multiple people through exchange omnibus accounts or multisig wallets.
Raw DAA counts systematically overstate real human participants while simultaneously undercounting users who interact exclusively through centralized exchanges, where activity occurs off-chain inside the exchange's internal ledger.
How the Metric Gets Distorted
The distortions fall into three categories. First, Sybil farming: individuals or coordinated groups create thousands of addresses to qualify for airdrops or token incentive programs. A single operator running scripted wallets can appear as thousands of organic users. Second, MEV and arbitrage bots: automated programs executing high-frequency transactions to capture value from transaction ordering account for a substantial share of on-chain activity, particularly on Ethereum. These are real economic actors, but they are not human users in any growth-metric sense. Third, dispersion contracts: smart contracts designed solely to receive funds and distribute them across many addresses, creating the appearance of broad activity from a single funding source.
A September 2025 analysis by a16z Crypto attempted to estimate real users by filtering out suspected bot and Sybil addresses using on-chain forensics. Their conclusion: across blockchains reporting roughly 220 million monthly active addresses, actual real monthly human users were likely between 30 and 60 million. That implies 70 to 85 percent of reported active addresses may not represent distinct human participants.
Companion Metrics That Tell a Better Story
An allocator evaluating blockchain adoption should pair DAA with at least three companion metrics. First, transaction value distribution: networks where median transaction value is rising alongside address counts show healthier adoption than those where growth comes from dust-level micro-transactions. Second, new address retention: of addresses appearing for the first time, how many remain active 30 or 90 days later? High new-address counts with low retention suggest incentive-driven or bot activity. Third, fee-paying addresses: addresses that pay meaningful gas fees demonstrate willingness to bear economic cost, which bots minimize. The ratio of fee-paying to total active addresses serves as a rough bot-filtering proxy.
From an incumbent's perspective, this is precisely the weakness a Visa or SWIFT executive would highlight. Traditional payment networks report verified unique users, KYC-backed transaction counts, and auditable merchant activity. Blockchain's pseudonymous architecture makes those guarantees structurally impossible without additional identity layers, which is why protocols like World Chain are building proof-of-personhood systems.
What Is Constructive About the Current Trajectory
The encouraging signal is that the industry is treating this measurement problem seriously. The a16z framework for estimating real users represents the first rigorous public attempt to quantify the gap between addresses and humans. On-chain analytics firms like Arkham Intelligence, Bubblemaps, and Nansen are building Sybil-detection tooling that protocols can integrate into incentive designs. And the emergence of AI-powered wallet agents, while it complicates the picture, is accelerating demand for on-chain identity standards that could eventually make address-level metrics more reliable.
For now, the practical takeaway is straightforward: any growth narrative built solely on DAA counts deserves skepticism, and the gap between reported addresses and estimated real users is the single most important adjustment an allocator can make when evaluating blockchain adoption.
For informational purposes only. Not an offer to buy or sell any security. Available only to accredited investors who meet regulatory requirements.
