Pick any "best Telegram group management bot" list and the features rhyme: welcome messages, word filters, anti-flood, a /ban command, maybe scheduled posts. All useful. And for years every one of those lists came with the same footnote — none of this stops a bot flood, so run a second bot for that. We wrote that footnote too. It is out of date now, and the reason is worth explaining properly.
What a management bot is genuinely good at
Credit where it's due. Combot, GroupHelp and Rose take the tedious, repetitive part of running a group off your hands: greeting new members and pinning rules, warn / mute / ban commands so co-admins aren't doing everything by hand, scheduled posts, basic activity stats. If your group is a hundred people who mostly behave, that toolset is the whole answer and you can stop reading.
The trouble starts when the problem stops being a person.
Two layers, and why one bot used to miss one of them
Every feature in that list reacts to a message. Someone posts, the bot checks the text against a rule, the bot acts. That loop assumes the threat says something. Coordinated accounts often don't — they join, they sit, they inflate your member count, and then a few hundred of them dump identical text in the same five seconds.
So there are really two layers. The message layer asks "should this text be here". The account layer asks "should this account be here at all". A word filter cannot flag an account that hasn't spoken, and an anti-flood rule counting messages per second does nothing about fifty accounts joining per second. Classic management bots own the first layer. TGuard was built for the second, which is exactly why the old advice was to run both.
Moderation acts on what a member says. Protection acts on how a member got in. Cover one and you are guessing about the other.
TGuard now runs the message layer too
The spam engine reads every message in a protected chat — and in the comments under channel posts, which is a separate chat Telegram calls the linked discussion group — and checks it against rules you set on a settings page for that specific chat.
- Stop words. Substring or whole-word match, your list, per chat. Whole-word keeps "cast" from catching "podcast"; substring catches the obfuscated variants people use to dodge simple filters.
- A link filter that covers
http://, bare domains, andt.me/telegra.phlinks. Off by default, because plenty of chats are built on people sharing links. - The known-abuser rule. If the sender is already flagged as an abuser across the network, their message is spam by definition — no keyword needed.
Each rule carries its own action: record it, delete the message, or delete and ban. New rules start in record-only mode on purpose. You watch what a rule would have caught for a day before you let it delete anything, because the fastest way to lose a healthy chat is a badly chosen stop word firing on real conversation. For the same reason there's a hard cap on how many times one rule can fire in one chat per window — if a rule goes berserk, it pauses itself and writes a warning instead of clearing the room.
How much traffic this is: in one ordinary 28-minute window on production, the engine logged 1,749 messages from 221 accounts already flagged as abusers, spread over 129 chats. Roughly a third of those — 561 — landed in the comments under channel posts rather than in group chats. Comments are where most owners aren't looking.
And the account layer, which hasn't changed
A captcha at entry. Every new member proves they're human before they can post. Cheap bought bots running a join script never clear it and never reach the chat.
A join-rate limit. Coordinated accounts arrive in bursts. Set a threshold — say, more than 10 joins in 30 seconds — and TGuard acts on the whole burst, ban or kick. The raid that would have dropped 40 accounts gets cleared before it lands.
A shared attacker database. When an account has been banned across five or more channels in the network, TGuard flags it as a known abuser and blocks it on sight — even an aged account with a real-looking profile that would sail through a captcha. That's bot protection based on documented history rather than a keyword guess, and it's the one piece no single-group bot can replicate: it needs the network.
So do you still need two bots?
For spam, no. If your management bot is there purely to filter junk, one of the two is now redundant, and the one that also sees join bursts and abuse history is the one worth keeping.
For everything else, quite possibly yes, and we'd rather say so than oversell. TGuard doesn't write welcome messages, doesn't schedule posts, doesn't hand you engagement charts, and won't referee an argument between two long-standing real members — that's a warn-and-mute job. If those matter to you, keep both bots. One rule if you do: give delete rights to TGuard only. Two bots racing to delete the same message produces confusing logs and the occasional message that survives because each bot assumed the other took it.
Setting it up
- Open @channel_guardian_bot and add it to your group as an admin with delete and ban rights.
- For comments under a channel's posts, add it to the linked discussion group — the channel and its comment chat are two different chats, and the bot has to be in the second one.
- Turn on Spam Protection, then open the settings page and add your stop words and link rule.
- Leave everything on record-only for a day. Read what it caught. Then switch the rules you trust to delete or ban.
- Set a join-rate limit that fits your normal growth, and if bot accounts are already inside, run a scan and clear them in bulk.
If you're comparing tools rather than adding one, the companion piece — choosing a Telegram anti-spam bot — walks through what the free options actually cover and where they run out.
Frequently Asked Questions
It automates the routine work of running a group: greeting new members, filtering banned words and links, muting or banning on command, scheduling posts, and logging admin actions. Combot, GroupHelp and Rose are the common ones. Where they stop is the account layer — a rule that reads messages cannot flag an account that has not posted yet.
Not since TGuard added message rules. It now runs stop words and a link filter on every message in the chat and in the comments under channel posts, and it also handles captcha, join-rate limits and known abusers. If your management bot exists only to filter spam, one of the two is redundant.
Yes, and for some setups that is still the right answer. TGuard does not write welcome messages, schedule posts or produce engagement stats. Keep the management bot for those, let TGuard own spam and account defence, and give only TGuard delete rights so two bots never race to delete the same message.