Checking whether a wallet qualifies for an airdrop is a read operation. It asks a question of public data: was this address holding, bridging or swapping when the snapshot was taken. That question can be answered from your public address alone, which is why a checker that insists on a wallet connection before it will show you a number has already told you something about itself.

Onchain Diary, in a breakdown of eligibility checker phishing, reduces it to five words: "A read never needs a write." That is the whole test. Keep it in your head and most of the fake checker problem stops being a judgment call and becomes a rule you can follow while distracted, which is the only kind of security rule that survives an airdrop week.

How a lookup turns into a write

The funnel runs in stages, and the first one looks harmless. A connect prompt gets you to authorize the site to see your address, and that is all it does: no funds move. Onchain Diary calls these Grade 1 wallet-connect harvesters, and notes what they are collecting is the address plus IP and session data, packaged into targeting lists for later attacks, usually behind a fake loading spinner and a believable "0 points" result.

Stage two is the one that costs money. Having connected, you are asked to sign something to "verify humanity" or "prove wallet ownership", and the signature is where write access lives. It might be an approve() call granting unlimited spending rights that get used later through transferFrom. It might be a Permit2 approval, which routes through a universal router and can reach several previously approved token balances from one prompt, something a genuine claim has no reason to ask for. It might be an EIP-2612 permit, an off-chain gasless approval that leaves no pending transaction in your activity history at all. Or, rarest and worst, a raw eth_sign blob presented as verification that can authorize an arbitrary transaction. CleanSky adds a newer one to the list: an EIP-7702 delegation that hands complete wallet control to a contract in a single signature, allowing every asset to be swept atomically.

The third stage exists too, and it is the crude one: a prompt to enter your seed phrase to "sync" or "import" the wallet so eligibility can be displayed. No application requires that, ever.

The money involved is not theoretical. ChainAware.ai's screener roundup puts approval-based attacks at over $200 million in losses across 2024 and 2025, sets the Inferno Drainer kit at roughly $80 million in 2023 alone, and cites around $800 million tracked by Scam Sniffer since 2023. Morten Christensen of AirdropAlert describes the mechanism without flattering the victim or the attacker: "A drainer does not 'hack' you. Instead, you authorize the theft yourself, usually without realizing what you just signed."

Three cases with the same shape

The FBI issued alert I-060325-PSA on June 3, 2025 about fake NFT airdrops on Hedera. Criminals used the network's legitimate memo field to embed a phishing URL in an unsolicited token. Click the link to claim, land on a page asking for a wallet connection or credentials, and the wallet is emptied to attacker-controlled addresses.

In March 2026 the same pattern ran on Tron as a fake "FBI Token" in TRC-20 form. The token's embedded message claimed the recipient was under an anti-money-laundering investigation and that assets would be frozen unless identity was verified at a linked site. Tronscan data cited in that reporting put the count at 728 wallets affected, several of them holding over $1 million in USDT each. The FBI's New York field office said plainly: "If you receive a token from an account with the details below, do not provide any identifying information to any website associated with such token."

The third is the clone. CLINKSINK, a Solana drainer, impersonated Phantom, DappRadar and BONK and took an estimated $900,000. That is the version most relevant to checkers, because a cloned tool with the right logo and a plausible domain is now cheap to produce. ChainAware.ai notes lookalike domains are often registered weeks or months before a project announces anything, and optimized so they surface first when search interest spikes.

Connecting is not signing

The fairest argument for connecting is a technically correct one: a connection by itself cannot move a token. It is an address disclosure, not an authorization. That is true, and it is why some people connect to anything without consequence for years.

What it ignores is that connecting buys you nothing a pasted address would not, while costing you the session and IP data that feeds the next attempt, and, per the same source, gradually training you to treat prompts from that site as routine. The habituation is the point of Grade 1.

Connect request Signature request
What the site gets Your address, plus IP and session data A cryptographic authorization tied to your keys
What it can move alone Nothing Up to unlimited token spend, or whole-wallet control via EIP-7702 delegation
Trace left in your wallet None An EIP-2612 permit leaves no pending transaction in activity history

The second confusion worth clearing: an eligibility check and a claim are different acts. The claim genuinely needs a write, because it is a transaction. But it should call a claim function, not an approval, a permit, or an account-delegation function. Cautellus states the default in its 2026 scam guide: "If it asks for unlimited or broad access rather than a specific amount for a specific transaction, treat it as dangerous by default."

The habit this should leave you with

Paste the address. If a checker will not return a result from a pasted public string, close the tab and find the number somewhere else. Which tools genuinely work read-only and which insist on a connection is the question our eligibility checker testing exists to answer, tool by tool, and it is worth knowing before a snapshot week rather than during one. The sources here do not settle it for every tool on the market; they settle the rule you apply to all of them.

When you do reach a real claim, spend twenty seconds on the prompt: read the function name, check the spender address in a block explorer, refuse anything with an unlimited allowance, and never sign a field you cannot see decoded. Then revoke. Revoke.cash has been doing approval audits since 2019 across more than 100 networks, and ChainAware.ai's advice is to clear approvals within 24 hours of any claim, not eventually.

And do it from a wallet you can afford to lose. Christensen's rule is the one I follow: "Keep a separate wallet with a small amount of funds for anything experimental. When you connect to a new site, you connect the burner, not the wallet holding your stack." That only works if the stack is actually somewhere else, on a separate seed, with the claim wallet parked near zero between seasons. His other line is the one that saves the most money: "Real opportunities survive you taking five minutes to verify. Scams do not."