An airdrop eligibility checker is a lookup. You give it a public wallet address, it compares that address against some version of a protocol's rules, and it tells you whether you're in. The phrase that matters is "some version." Every checker you'll paste an address into reads one of three quite different sources, and the source decides whether its yes means anything.

That is why two checkers can look at the same wallet on the same afternoon and give different answers. It is also why neither of them should ever need more from you than the address.

Three sources, three levels of trust

The protocol's own list. The first kind of checker reads an eligibility list the protocol computed itself. Wormhole's is the clearest documented case. Eligibility for the W drop was checked at airdrop.wormhole.com by entering an address or connecting Discord, and the list behind it came from a pipeline no outsider could reproduce. Wormhole's own write-up describes an "ownership graph" that assigned one user ID to a cluster of wallets. It weighted wallets by volume, timing and consistency, and it set a minimum cumulative value of roughly $1,500 after adjustments. Only this kind of checker knows the final answer, and even it can show you a stale one: Wormhole told users to hard-refresh if a cached result looked out of date.

A rebuild from the announced rules. The second kind takes whatever criteria a protocol has published and replays your on-chain history against them. Guava looks at transaction counts, bridge activity, wallet age, DEX volume and governance votes, and it says plainly that its criteria come from "known snapshot requirements and community-researched farming guides." AirdropAlert's Dropsbot engine does the same across more than 360 tracked campaigns. A rebuild is good at facts, such as whether you bridged or held through a date. It is weak on anything the protocol scores privately. Wormhole gave timing bonuses for first-year mainnet use, for three or more months of consistent activity, and for activity in the June 2022 to October 2023 bear market. It also applied a negative multiplier for thin-liquidity assets. A rebuilder can check most of that. It cannot see the sybil review that comes afterward.

An estimate before any list exists. The third kind is a guess. Passive Yield Lab's checker doesn't scan your wallet at all. You tick activities off a checklist and score up to 120 points, with 15 for bridging to an L2. Points dashboards belong in this category too, because they count activity before the protocol has decided what the activity is worth.

Once you know the three sources, the disagreements stop being mysterious. They come down to three things:

  • Which source the checker uses. A list, a rebuild and a guess will not agree with each other.
  • When it last updated. Wormhole's March 11 update added community cohorts and improved Sui and Algorand wallet compatibility. Any checker that copied the rules before that date was already wrong.
  • Which chains it indexes. Wormhole reconciled activity across 30 blockchains, while Guava supports seven named protocols. If your qualifying transaction sits on a chain the checker doesn't read, it will report nothing.

None of this requires your keys. Wallet history is public, so reading it needs only the address, and AirdropAlert's checker is built to take a pasted address with no wallet connection. A signature is a different thing entirely: it grants permission. When a "checker" asks you to sign, it is asking to act on your behalf, not to look something up. The difference between a read-only lookup and a signature request is covered in full in is it safe to connect.

Where the gap showed up

Linea is the case to remember. According to The Block, Linea and Nansen screened 1,297,203 addresses that held Linea XP and flagged 516,960 of them as sybil, leaving 780,243 eligible. Any checker working from XP data would have told about four in ten of those wallets they were in when they weren't.

Optimism shows the same gap arriving later in the process. It excluded roughly 17,000 addresses after its eligibility announcement and recovered about 14 million OP. A checker that was accurate on announcement day was wrong a few weeks afterward.

Wormhole shows the gap working in the other direction. Its community allocation included the top 500 Zealy contributors, the top 10,000 Pyth stakers, several NFT communities and 1,081 Monad community members. Some of those qualified through Discord rather than through their on-chain activity. A rebuild that reads only on-chain data would have told some of them no.

The fairest case for third-party tools is coverage, and it is a good one. DefiLlama's checker takes one EVM or Solana address, reports which tracked drops it qualifies for, and says it queries only DefiLlama's own APIs. Nobody is going to visit 360 claim sites by hand. That argues for using aggregators to find out where to look. It does not argue for trusting them on the answer, because none of them can see the sybil sweep that hasn't happened yet.

Not a claim page, not a snapshot

Eligibility checker Often confused with
Claim page Reads a public address and returns a result. No signature needed. Moves tokens, so it needs a wallet connection. Fake claim pages frame spend approvals as "verification."
Snapshot Reports what a source believes about your wallet. The recorded moment itself. Activity after the snapshot doesn't count, whatever a checker says.

The first confusion is the one that gets wallets drained. On August 6, 2026, the XRP Ledger Foundation and RippleX warned about an "XRP Rewards Scanner" that promised to find unclaimed XRP if you connected your wallet. RippleX said plainly that no such scanner exists. A checker that wants a connection is behaving like a claim page, and fake claim pages are where drainers operate.

Three questions before you believe a yes

Ask every checker the same three things. Whose list is it reading: the protocol's, a rebuild, or a guess? When did it last update against the protocol's rules? Does it index the chain where you actually did the work?

Until the protocol's own checker is live on its own domain, treat any yes as provisional. Before you worry about a no, confirm you pasted the address you actually used. A custodial exchange address, a typo or the wrong network format will produce a false answer from even the best tool. Then check that address against your own history on a block explorer.