Fake Support Agents In Your Direct Messages: How The Impersonation Script Works
The message arrives before the answer does
You’re stuck. A claim page shows zero, a bridge transfer hasn’t landed, or a points dashboard says your wallet isn’t eligible when you’re fairly sure it should be. So you do the sensible thing and ask in the project’s Discord or Telegram group. You post in the support channel, describe the problem, maybe mention which wallet you use.
Within a few minutes, you get a direct message. The account has the project’s logo as its avatar. Its name says something like “Support” or “Mod” next to the project name. The message is polite and specific: it saw your question in the support channel, it’s sorry for the trouble, and it can help you fix it right now.
That speed is the first tell, and most people miss it because it feels like good service. Real project teams are usually slow to answer in public channels, let alone privately. The fake agent is fast because it’s watching for exactly the kind of message you just posted.
This piece walks through how that script runs from the first DM to the moment funds leave, why legitimate teams structure their support so that it never starts in your inbox, and the settings on Discord and Telegram that stop most of these messages from reaching you at all. It also covers why farmers, who ask a lot of questions across a lot of communities, are a particularly good audience for it.
None of this requires a technical mistake on your part. There’s no dodgy contract you forgot to check. The whole attack runs on trust, speed and the relief of finally getting help.
How the attackers find you
Fake support operations don’t pick victims at random. They watch public channels for signals that someone is stuck and has funds.
The simplest method is keyword monitoring. Accounts sit in popular project servers and groups, often many at once, and flag messages containing words like help, stuck, claim, error, wallet, bridge, pending, missing or eligible. A message like “my claim shows zero, can someone help” is ideal: it says the author has a wallet, expects tokens, and is frustrated enough to act quickly.
Timing helps them. The busiest moments in any project community are launches, claim windows and outages, when many users are confused at once and the real team is overwhelmed. Those are the hours when real answers are slowest and fake ones are fastest.
Member lists give them reach. On Discord, anyone sharing a server with you can see your profile and, depending on your settings, message you directly. On Telegram, anyone in a group with you can usually see your name and, if you have one, your username, which is enough to start a private chat. Neither platform requires the sender to know you.
Your own posts fill in the details. If you mention which wallet you use, which chain, or paste a transaction hash or address, the attacker now knows how much you’re holding and what story to tell. An address in a public channel is an invitation to look you up on an explorer.
And once your account has responded to one fake agent, it tends to end up on lists of accounts worth trying again.
The script, step by step
The details vary, but the structure is remarkably consistent. It follows the shape of any good sales conversation, which is part of why it works.
Step one is rapport. The agent references your exact question, which proves they read it and makes them feel connected to the project. They’re friendly, patient and seem competent.
Step two is authority. They might give you a ticket number, use the project’s terminology correctly, mention that “the dev team is aware of the issue”, or send a screenshot of an internal-looking dashboard. Some accounts copy a real moderator’s display name and avatar exactly.
Step three is diagnosis. They explain your problem in a way that sounds plausible: your wallet needs to be “synced” with the new contract, your node is out of date, your address wasn’t “validated” before the snapshot, or there’s a mismatch between your wallet and the claim portal.
Step four is the fix. This is where the attack lives. They send you a link to a site that will “rectify”, “sync”, “validate” or “migrate” your wallet. Or they invite you to a “support server” with a ticket bot. Or they ask you to fill in a form. Or they ask to screen share.
Step five is urgency. If you hesitate, the tone shifts. The problem will get worse, your allocation will be forfeited, the window closes in an hour, your wallet will be flagged. Sometimes a second, more senior “agent” joins to add pressure.
Step six is the take. The site asks for your seed phrase, or a signature, or an approval, and the funds go.
The words that should end the conversation
A handful of phrases come up again and again in these scripts, because they sound technical enough to be credible and vague enough to fit any problem.
“Sync your wallet.” Wallets don’t need syncing with a project. A wallet is a key and an interface that reads the chain. There’s nothing on your side for a project to sync with.
“Validate” or “verify” your wallet. No legitimate claim, bridge or points system asks you to validate a wallet on a separate site. Eligibility is a lookup against your public address. It doesn’t need anything from you except, at most, the address.
“Rectify”, “migrate” or “restore”. These show up when the story is that something broke. A real migration, where a project moves to a new contract, is announced in public by the project, applies to everyone and happens through the official app.
“Connect to the dapp to resolve the issue.” Connecting a wallet to a site is how you begin interacting with that site, which is exactly what the attacker wants you to do. No support procedure starts there.
“Enter your recovery phrase to confirm ownership.” Nobody ever needs your seed phrase. Not support, not a validator, not a form. A seed phrase is total control of every account derived from it. Anyone asking for it is asking for the wallet.
If any of these appear in a message you didn’t start, you can end the conversation there without needing to work out whether the sender is real. The request itself answers the question.
Why real teams don’t DM first
It’s worth understanding why legitimate projects run support the way they do, because it makes the fake version easier to spot.
Real teams handle support in public or in structured systems. That usually means a ticket system inside the official server, where you open a ticket and a private channel appears between you and staff, visible to the server’s moderators. Or it means a help centre, a support email on the official domain, or answers in the public support channel. Each of these keeps a record the team controls.
Direct messages don’t fit that model. The team can’t see or audit what a moderator says in a DM. They can’t hand the conversation to someone else. They can’t stop a compromised moderator account from saying something harmful. So most serious projects tell their staff not to initiate DMs, and tell their users, often in a pinned message or the server rules, that the team will never DM first.
There’s also a practical reason. Real support usually can’t fix your problem from a DM anyway. If a claim shows zero, the answer is almost always either that your address wasn’t in the eligibility list, which support can’t change, or that there’s a bug in the official app, which gets fixed for everyone. Neither involves you visiting a separate site.
So the rule is simple and holds up well: anyone who DMs you first offering support is not support. You don’t need to verify the account to act on that.
What a real support interaction looks like
It helps to hold a picture of the genuine version in your head, so the fake one feels wrong by contrast.
A real answer usually comes in public, in the channel where you asked, and it’s often slower than you’d like. A moderator with a visible role replies in the thread, sometimes hours later during a busy launch. The reply tends to be generic: there’s a known issue with the claim page and the team is working on it, or the eligibility list is final and can’t be changed, or here’s the documentation page that explains the criteria. Real support is frequently unsatisfying because it can’t make your wallet eligible if it wasn’t.
When a project runs private tickets, you start them. You go to the official server, find the ticket channel listed in the server’s own navigation, press a button, and a private channel opens that the project’s staff can see. The staff member in that ticket asks about symptoms. They might ask for your public address so they can look it up. They won’t send you to a different website to fix anything, because the fix, if there is one, happens on their side or through the official app.
Real support is also comfortable with “we don’t know yet”. During an outage or a contested snapshot, the honest answer is often that the team is investigating. Fake support always knows exactly what’s wrong and exactly how to fix it, within minutes, which is itself suspicious.
And real support never asks for anything that grants control. It can ask what you see. It can ask for a transaction hash. It can’t ask you to sign, approve, install, share your screen or pay a fee. Once you’ve seen a few genuine interactions, the fake ones stand out because they’re too helpful, too fast and too private.
Why checking the account isn’t enough
People often try to work out whether a DM is genuine by checking the sender’s profile. It feels like diligence. It’s a weak test.
Display names and avatars are free to copy. On Discord, the display name that shows in chat can be anything, including a real moderator’s name, and the avatar can be the same image. Usernames are unique, but most people never look at them. On Telegram, display names are similarly free, and a username can differ from the real admin’s by one character.
Mutual servers prove nothing. The fake account is in the same server as you because that’s where it found you.
Roles are a better signal but still imperfect. On Discord, a real moderator has a role visible in the server’s member list, and you can check whether the account messaging you is the same account listed there. That catches most copycats. It won’t catch a genuine moderator account that’s been compromised, and that happens: moderators get phished too, and a real account with a real role sending a malicious link is the most dangerous version of this attack.
Account age helps a little. Many fake accounts are new, though attackers also buy or hijack aged ones.
This is why the behaviour-based rule beats the identity-based one. You can spend ten minutes investigating whether an account is really a moderator and still be wrong. You can spend two seconds noticing that it DM’d you first and asked you to visit a link, and be right.
The fake ticket server
A more elaborate variant moves the conversation out of DMs, which makes it look more legitimate.
The agent tells you that your issue needs a support ticket, then sends you an invite to “the support server”. The server looks like an official one: the project’s branding, channels with familiar names, a ticket bot that greets you with a menu. You open a ticket, and a “staff member” appears in a private channel to walk you through the fix.
Everything in that server is controlled by the attacker. The bot, the staff, the welcome messages, even a crowd of other “users” who seem to have had their issues solved. The ticket process adds a sense of structure that makes the eventual request feel routine.
The defence is to only open tickets in the project’s official server, reached through a link on the project’s official website. If a ticket system exists, it’ll be there. An invite link sent by someone who messaged you first is never the path to a real ticket.
A related trick shows up on Telegram: cloned groups. A group with the project’s name and logo, sometimes with more members than the real one, runs the same scripts. Many projects list their official Telegram links on their site. Anything else is someone else’s group.
Verification bots and the paste-this trick
There’s another pattern that’s become common in crypto Telegram groups, and it’s worth knowing because it doesn’t look like support at all.
You click a link to join a group and land in a channel with a “verification” bot. The bot asks you to prove you’re human. The steps look like a captcha, then it asks you to copy a command, open a system dialog on your computer, paste the command and press enter. The instructions are phrased as a routine anti-spam check.
That command downloads and runs malware. Often it’s the kind that steals browser data, including extension storage, saved sessions and anything in the clipboard, or that replaces addresses you copy with the attacker’s. You haven’t signed anything. You’ve installed something that can wait for your next transaction.
No legitimate verification ever asks you to paste a command into your operating system. Captcha bots in real groups work inside the chat app, usually by pressing a button or solving a simple puzzle. If a group’s entry process involves your computer’s run dialog or terminal, leave.
This matters especially on a machine where you sign transactions. A separate piece covers malicious extensions and clipboard hijackers in more detail, but the short version is that the chat app you use for communities shouldn’t be on the same machine that holds your signing wallets, if you can arrange it.
Screen sharing and remote access
Some scripts skip the malicious site and ask to see your screen instead. The agent says it’s easier to fix the problem if they can see what you see, and suggests a screen-share call or a remote desktop app.
Screen sharing is dangerous even without remote control. If you open your wallet while sharing, the agent sees your balances and addresses. If they can talk you into revealing your seed phrase “to check it’s the right wallet”, it’s now in a recording. If a wallet shows a secret recovery phrase screen at any point during the session, it’s gone.
Remote access is worse. A remote desktop tool gives the other person control of your mouse and keyboard. With your wallet unlocked, they can approve transactions themselves. With it locked, they can install something that waits for you to unlock it later.
There’s no legitimate crypto support flow that requires seeing your screen. Projects can’t see your wallet from their side, and they don’t need to see it from yours. If a conversation gets to “can you share your screen”, it’s already gone wrong.
Why farmers are a good audience for this
Anyone in a crypto community can be targeted, but people farming airdrops tend to tick more of the boxes attackers watch for.
They ask a lot of questions. Farming means dealing with testnets that break, eligibility checkers that return confusing results, points dashboards that lag and claim windows that open without warning. Each confusion is a reason to post in a support channel.
They’re in a lot of communities at once. A farmer following fifteen projects is visible in fifteen member lists, and every one of those servers can host a fake agent who can see them.
They often ask about eligibility, which is exactly what the script is built around. “Why does my wallet show not eligible?” gives the attacker a ready-made story: your wallet wasn’t validated, and here’s where you validate it.
They sometimes post addresses. It’s tempting to paste your address to prove a problem, and that ties your chat account to your wallet in public. Beyond the scam risk, that’s an off-chain link between a social account and an address. If the same chat account posts several addresses over time, it has publicly connected them all.
And they may run several chat accounts, which multiplies the surface. Each account needs its own settings, and it’s easy for one to be left on defaults.
One conversation, start to finish
Here’s a composite of how a typical exchange goes, built from the common steps rather than from any single real case.
A user posts in a project’s support channel: “Bridged to the new chain two hours ago and the funds still haven’t arrived. Anyone else?” Four minutes later, a DM arrives from an account called “Support Team” with the project’s logo.
The agent opens warmly. It saw the question, there’s a known delay affecting some users, and it can help. It asks which wallet the user has. The user says which one. The agent replies that this wallet is affected by a “sync issue” after the recent upgrade, and that transactions stuck in the bridge need to be “rectified” through the recovery portal. It sends a link. The domain includes the project’s name and the word “recovery”.
The user hesitates and asks why this isn’t in the announcements. The agent says the team doesn’t want to cause panic, so it’s contacting affected users individually. It adds that pending transfers expire after twenty-four hours if they’re not rectified.
The user opens the link. The page looks like the project’s app, asks to connect a wallet, then shows a “rectify transaction” button. The wallet pops up a signature request. The user signs. Nothing visible happens. The agent says it can take a few minutes to process. Twenty minutes later, the wallet’s stablecoins move to an unknown address through a permit signature the user just gave.
The original bridge transfer, meanwhile, arrives an hour later. It was just slow.
Every stage of that conversation had an exit. The DM arriving first was one. The link to a different site was another. The explanation for why it wasn’t public, the deadline, and the signature request were all exits too. The script only works if you take none of them.
The settings that cut exposure
You can’t stop scammers existing. You can make it harder for them to reach you.
On Discord, the most useful setting is per server: turn off direct messages from server members. You’ll find it in the server’s privacy settings. You can also set a default for new servers you join, so DMs from members are off unless you choose otherwise. Discord also has message request and spam filtering options that send DMs from people you don’t know into a separate area. Turn those on.
Consider friend request settings too. Limiting friend requests to friends of friends or members of specific servers reduces another route in.
On Telegram, check who can see your phone number and who can find you by it, and set both to nobody or contacts. Check who can add you to groups, since being added to a cloned group is a common starting point. Telegram has also added controls over who can start new chats with you, with some options tied to its paid tier. Use whatever’s available on your account.
Both platforms let you hide or limit your profile details. A plain username that doesn’t reference crypto, wallets or airdrops is slightly less interesting to someone scanning a member list.
These settings won’t stop someone who’s determined, and they don’t replace the rule about unsolicited support. They just reduce how many of these conversations start.
How to ask for help without inviting a scammer
You’ll still need help sometimes. There are ways to ask that make you a worse target.
Start with the official documentation and announcements. Many “is this broken?” questions are answered in a pinned message or a status post. If the claim page is having problems, the team usually says so in the announcements channel.
When you do post, don’t include your address, balances or wallet name unless a real staff member asks in public. Describe the symptom: “The claim page shows zero for an address I expected to be eligible. Is there a known issue?” That’s enough for real support to respond and gives an attacker less to work with.
Say, in your message, that you won’t respond to DMs. It sounds odd, but it tells anyone watching that you know the script, and it reminds other readers too.
Use the official ticket system if one exists, reached from the project’s site, not from a link someone posted.
And if you get a DM anyway, don’t reply. Replying confirms the account is active. Report it through the platform, block it, and if the server has a channel for reporting scammers, post the account’s username there so the moderators can remove it.
If you already clicked or signed something
If you’ve followed a link from a fake agent and connected a wallet, the next steps depend on what you did on that site.
If you only connected and didn’t sign or approve anything, the risk is mostly that the site has your address and knows you’re susceptible. Close it, don’t return, and treat any follow-up contact as hostile.
If you signed a message or approved a contract, act quickly. Check your approvals with a revoke tool and remove anything granted to the site’s contracts. Be aware that some signatures, like off-chain permits, can be used without leaving an approval you can see until they’re executed. If there’s meaningful value in the wallet, move it to a different wallet you control and trust, one whose seed has never been anywhere near the incident. The separate piece on recovering after signing a drainer approval goes into this in more detail.
If you entered your seed phrase anywhere, assume every account derived from it is compromised. Move funds from all of them to a new wallet with a fresh seed, as fast as you can, starting with the largest balances.
If you ran a command or installed software, treat the machine as compromised. Don’t use it for wallet activity until it’s been cleaned properly.
Expect a second wave. People who’ve been scammed get targeted by “recovery” services offering to get funds back for a fee. They’re the same kind of operation.
What alternatives get wrong
A few common pieces of advice are well meant but don’t hold up.
“Only trust accounts with the moderator role.” Better than nothing, but compromised moderator accounts exist, and role colours can be imitated in screenshots. The rule about DMs and links works regardless of role.
“Check the link before clicking.” Checking the domain is useful, but fake support scripts often use domains that look plausible, including ones that include the project’s name. A link sent in a DM you didn’t ask for is already disqualified by where it came from, so checking it is wasted effort.
“Use a burner wallet with the support site.” Connecting a burner to a malicious site isn’t a clever workaround. It wastes time, still exposes a device and browser to a hostile party, and can create links between the burner and your other wallets depending on how you funded it.
“I’m too experienced to fall for this.” Experience helps with the obvious versions. It doesn’t help much at two in the morning during a claim window, after three hours of a broken dashboard, when a friendly person offers to fix it in five minutes. The script is designed for that moment, not for your best day.
Where the rule has limits
The rule that support never DMs first catches nearly all of these scams, but it’s worth being honest about what it doesn’t cover.
It doesn’t protect you from a malicious link posted publicly by a compromised official account. That happens when a project’s Discord bot or a moderator account gets taken over and posts a fake announcement in the official channel. The defence there is to cross-check any urgent announcement against the project’s website and other official accounts before acting, and to be suspicious of any announcement pushing you to act within minutes.
It doesn’t protect you if you start the DM yourself with someone who turns out to be fake. If you message an “admin” you found in the member list, you’ve skipped the tell. Use the official ticket system instead.
And it doesn’t stop legitimate communities from occasionally using DMs for harmless things, like a moderator following up on a report you filed. The line to hold is what’s being asked for. Nothing legitimate needs your seed phrase, a signature on a site linked in a DM, remote access or a payment.
I don’t have reliable figures on how much is lost to fake support scams compared with other routes, and I’d treat any widely circulated number with caution unless its method is shown. What the mechanics show clearly is that the script depends on you acting fast in a private conversation, and both of those are things you can simply decline to do.
For more on wallet security, community research and the habits that keep a farm running without handing it to strangers, head back to the Airdrop Farming homepage.
Get new guides and videos first — join the Telegram channel.