How to send bulk DMs on Telegram without getting banned

How Telegram restrictions actually work

There is no single "ban". There are three different states, and confusing them is why people keep applying the wrong fix.

PeerFlood is a rate limit. The API starts refusing new conversations with people who are not in your contacts. Your account still works normally — you can read, post and reply to existing chats. It usually clears on its own within hours to a couple of days. Nothing is lost; you were simply moving too fast for a session that young.

SpamBlock is a reputation penalty, and it is the one that matters. It is driven mostly by reports, not volume. Recipients pressed "Report spam", and Telegram limited your ability to message non-contacts. Duration ranges from a few days to permanent depending on history. You can ask @SpamBot for status and file an appeal, but outcomes are inconsistent.

Account deletion is terminal. The number is blacklisted and everything — chats, channels, contacts — is gone. This is what repeated SpamBlocks eventually turn into.

The thresholds that hold in 2026

Telegram does not publish limits and they are not fixed numbers — they scale with account age and trust. These are the working ranges we see across the platform:

• A fresh account: 2–3 messages to strangers before the first restriction. In practice, treat a brand-new account as unusable for outreach.
• After 2–4 weeks of aging plus warmup: 20–40 DMs per day per account, spread across the day, not fired in a burst.
• Group invites: 3 per day in week one, 5 in week two, 7 in week three, then 9–10. Going straight to 50 is the single most common way to burn an account.
• Daily account loss above 5% means the campaign is broken. Stop, look at the message, do not "push through it".

Note what these numbers imply: throughput comes from the number of accounts, not from pushing any single one harder. Ten warmed accounts at 30 DMs each beat one account at 300 every single time — the first setup survives, the second dies on day one.

Warmup: the part everyone skips

A warmed account is one whose history looks like a person's. Before it sends a single cold message it should have joined a few channels, read them, reacted to posts, held short conversations, set an avatar and a bio, and done all of that spread over days rather than in one scripted hour.

Two details do most of the work. First, let the account rest 24–48 hours after changing profile data — a name and avatar change immediately followed by outbound messages is a strong bot signal. Second, vary the timing. Perfectly uniform intervals are detectable by timing analysis alone, regardless of how human the text is.

Accounts that go through a proper warmup cycle last roughly an order of magnitude longer than accounts that do not. That is not a marketing figure — it is the difference between replacing your account pool every week and replacing it every quarter.

One account, one IP — and why it is not negotiable

Telegram links accounts that share a fingerprint or an address. If five accounts sit behind one IP and one of them gets reported, you do not lose one account — you lose the cluster. This is the mechanism behind "I woke up and everything was banned".

The rule is one account, one dedicated IP. Residential and mobile proxies work best because they belong to address ranges shared with real subscribers, which makes them expensive to blanket-ban. Datacenter IPs are cheaper and correspondingly easier to identify.

Geography matters too: an account registered on a US number that suddenly appears from a datacenter in another hemisphere is a cheap signal to catch. Keep the proxy region consistent with the account's origin, and keep it stable — rotating a session's IP every few minutes looks worse than not rotating at all.

Your message is the actual ban risk

SpamBlock is report-driven. That makes the text of your first message the highest-leverage variable in the entire operation, and the one people spend the least time on.

Identical text sent to thousands of people is trivially clustered, so uniqueness matters — but naive spintax produces sentences no human would write, which gets reported anyway. Rephrasing each message with a model, keeping the meaning and varying the structure, holds up better.

What reduces reports more than uniqueness, though, is relevance. A message that opens with a question about something the recipient actually posted gets answered. A message that opens with your offer gets reported. If your reply rate is under 2%, you have a copy problem, and no amount of proxy hygiene will fix it.

What to do when an account gets limited

1. Stop that account immediately. Continuing to send from a limited session converts a temporary restriction into a permanent one.
2. Ask @SpamBot for status. It tells you whether the limit is temporary and, roughly, when it lifts.
3. Do not touch the profile. Changing name, avatar or bio while restricted reads as evasion.
4. Let it sit and behave normally. Reading channels and replying inside existing chats helps; new outbound does not.
5. Fix the campaign, not the account. If several accounts got limited on the same script, the script is what is broken.

How this is handled in DarkTelegram

Everything above is operational discipline, and discipline is exactly what breaks when you are running dozens of accounts by hand. The platform enforces it instead:

• Conservative default limits per account, with randomised pauses between actions rather than fixed intervals.
• Staged warmup that runs in the cloud on its own schedule, including the rest period after profile changes.
• A Trust Score on every account, so you can see which sessions are healthy before you load them into a campaign.
• One dedicated proxy per account, from a built-in residential pool or your own.
• AI rewriting of every outgoing message, so recipients do not receive identical text.

None of it removes ban risk — nothing does, and anyone claiming otherwise is selling something. It removes the failure modes that come from doing this manually at scale.