How account protection works

Instagram does not usually restrict an account for sending too much. It restricts it for sending in a pattern — a burst of near-identical messages, usually right after a post takes off. Every control below targets the pattern rather than the volume.

Send caps, well under what Meta allows

Meta publishes a ceiling of 750 private replies per hour, per account. hellodm sends at 600 per hour — 80% of it — for every account, paced to no more than 10 a minute so it never arrives as a burst.

That headroom is deliberate, and so is what it is measured against. 750 is the point where Instagram’s API starts rejectingrequests. It is not the threshold Instagram’s spam classifier reacts to, and those are two different systems: an account can stay comfortably under 750 and still be actioned for pattern rather than volume. Meta publishes nothing about the classifier, so headroom against the number it does publish is the only kind anyone can reason about honestly.

The cap does not vary by follower count. It used to, and that was wrong: a reel from a 200-follower account can take off exactly like one from a large account, and a cap tied to audience size throttled it hardest at the moment its automation mattered most.

Meta also reports how much of your real budget is left on every API response. hellodm reads that header and backs off on it — that signal is authoritative, and the numbers above are the floor beneath it.

Surge cooldown

If more than 120 comments or messages arrive within five minutes — a post genuinely going viral, more than twice as fast as replies can go out — sending pauses for thirty minutes. Nothing is thrown away: the messages queue and are drained gradually once the cooldown ends, up to ten per account a minute. They are never released in a burst, which would recreate the exact pattern the cooldown existed to prevent.

Messages that would outlive Meta’s own reply window expire instead of sending late, so a backlog cannot turn into a wave of replies to week-old comments.

Message rotation — and what it is not

You can write up to four versions of a reply. Each send picks one of them at random, so a hundred people who comment do not all receive a byte-identical DM. Identical repeated text is the clearest signal a spam classifier keys on, which is why this is available on every paid plan rather than sold as a premium add-on.

To be exact about it: no AI model rewrites or generates your messages. Rotation chooses between the versions you wrote yourself, and nothing else. Your DMs say what you typed, every time — which also means the variety of your messages is only as good as the versions you write. Four genuinely different wordings protect an account; four near-copies of the same sentence do not.

Checks that run before a rule can go live

These are blocks, not warnings — a rule that fails one cannot be activated:

• No link in the first automated message. The single most-cited trigger in real account restrictions. Send the link as a follow-up once the person has replied.
• Opt-out text is required on every automated thread — Meta’s DM automation policy asks for it explicitly.
• No near-duplicate of another live rule on the same account and trigger. Two rules sending near-identical text is the same pattern problem as one rule with no variants.

What you control

Send pacing and surge protection can both be switched off per account in automation settings. They are on by default and we recommend leaving them on — with pacing off, sends go out as fast as Meta itself allows, bounded by nothing else.

The account-health screen shows your current cap, what has been sent this hour, what is queued and what is retrying. It is available on every plan, including free.

What this cannot promise

No tool can guarantee an account is never restricted — Meta does not publish the classifier, and its thresholds move. What these controls do is keep your sending well inside the range where accounts are not flagged, and refuse the specific patterns that show up again and again in real restriction reports. Anyone promising more than that is guessing.