Many of us have participated in airdrops and retrodrops, some received rewards, some didn’t. Projects often base their final selections on metrics like user onchain balance, address age, bridge volume into L2s, some txid like activity in dapps, and activity across multiple networks. But an increasingly tough challenge is: how do we distinguish real users from bots or farmed addresses? Especially when some farms operate manually, mimicking real user behavior.
In your opinion, what criteria separate a real wallet from a bot/farm account?
What attributes do you consider most important?
How would you filter out these artificial addresses?
This is a really tricky question, and realistically in my opinion, it’s not possible to track and separate every single address especially when AI agents are able to closely mimic real user behavior.
However, certain criteria can help reduce bot and farm activity. For example:
• Diverse, natural activity over time
• Organic social ties, such as interactions with multiple protocols, communities, or social verifications
• Reasonable timing patterns that don’t show large bursts of coordinated claims
• Holding different tokens and NFTs, which often indicates a real user.
I would prioritize wallets that engage with the protocol beyond rewards (e.g., governance participation, content creation), provide liquidity or sustained contributions and show longer-term, varied on-chain histories.
In some cases, optional identity or social verification can be considered, provided it doesn’t compromise decentralization principles.
With all said, it’s very difficult to prevent farming completely, but by applying these filters and monitoring behavioral patterns, we can at least avoid massive bot attacks and reduce the scale of fake account farming.
About AI agents for farming: for every gangster there is a policeman, and as soon as AIs appear that create such activity, a special police AI will appear that will see traces of the use of AI or simply find out such bot in public database of AI agents