AI Crypto Project User Growth: How to Filter Out Bot Activity

When examining user growth data for AI crypto projects, you must first assume the presence of large numbers of bots, then filter them out. This is not prejudice but industry reality. The on-chain identity system Matchain has already covered over 1.5 million users, of which 150,000 have been tagged as real identities through AI verification, suggesting that the real person ratio may be only one in ten. Sam Altman has also pointed out that the number of LLM-driven accounts on X platform surged in 2025, forcing crypto traders to reassess the reliability of social sentiment indicators.

Prerequisites

• Ability to view metrics like "total users" and "daily active users" published by the project.
• Access to on-chain data platforms (e.g., Dune, Nansen) or the project's own public dashboard.
• Knowledge of whether the project has "airdrop expectations" — the biggest driver of bot activity.

Step 1: Directly Check if the Project Published "Real User" Filtered Data

This is the fastest way. Some projects have already done this work.

How to do it: Search the project's official documentation or blog for keywords like "Sybil," "bot detection," "human verification." If the project itself has released a real user count after removing bots, adopt it directly.

For example: Matchain's MatchID module uses AI to analyze user behavior — interaction frequency, operation patterns, on-chain activity — to determine whether a user is real, and has already labeled 150,000 human identities. Galxe explicitly states that its bot identification system marks suspicious accounts through behavioral analysis and on-chain data analysis, and identified bots are disqualified from rewards.

Completion criteria: You have confirmed whether the project has released bot-filtered user data.

Step 2: Reverse-Engineer the Real Ratio Using Behavioral Metrics

If the project hasn't published filtered data, examine behavioral data yourself.

How to do it: Focus on the following metrics:

• Active time distribution: Real people usually follow circadian rhythms, concentrated in a few time zones. If user activity is evenly distributed throughout the day, the bot ratio is high.
• Operation intervals: Real people have thinking time when completing tasks; intervals are irregular and longer. If all "users" complete the same series of operations within seconds, they are likely scripts.
• Task completion rate: If the number of people completing one task is almost identical to the number completing the next task, it means the same batch of bots is running through the entire workflow.

Completion criteria: You have visually judged whether the project's user behavior is "human-like" or "bot-like."

Step 3: Check for "Proof of Personhood" Mechanisms

Projects with the strongest bot resistance often embed "Proof of Personhood" mechanisms. World's AgentKit allows AI agents to carry cryptographic proofs indicating they are backed by unique individuals verified through World ID, which can be used to limit daily usage per person regardless of how many agents are deployed.

How to do it: Check if the project documentation mentions the following keywords: World ID, proof of personhood, biometric verification, social identity binding. If none exist, the bot filtering capability is basically zero.

Completion criteria: You have assessed whether the project has the technical foundation to resist Sybil attacks.

Common failure reason: Looking at the "total registered users" figure published by the project and assuming it represents active users. The data of most airdrop projects includes massive numbers of "zombie accounts." The truly effective metric is "the number of users who continue to stay and generate transactions after KYC or behavioral verification," a figure that projects usually do not proactively disclose.

Risk warning: Do not decide to invest based solely on "user growth" data. If 90% of users are bots, the project's real ecosystem scale may be only a fraction of the advertised number. Research in 2025 has shown that Twitter sentiment correlates with BTC price, but an increased bot ratio severely distorts quantitative signals, and the same applies to on-chain user data. Regulators have reminded investors not to make trading decisions based solely on social media hype.

Verification Method After Completion

Search for the two metrics "daily active users" and "unique wallets" on Dune or the project dashboard. If daily actives are far lower than total wallets, it means most wallets are dormant. Divide active wallets by total users to get the "true retention rate" — a figure below 10% is normal, below 5% is a danger signal. Verification channels: project official dashboard, Dune, Nansen.