Getting a damaged sending domain back
Sooner or later a domain gets into trouble: a burst of cold mail, a bad address list, a loop nobody caught for six hours. Placement drops, and the instinct is to fix the content and keep sending, which is the one response that reliably makes it worse.
4 min read
Recovery is slow, boring, and entirely doable if you diagnose before you act.
Stop sending the thing that caused it
Before anything else, halt the traffic that did the damage. Not all sending, just the category responsible, which is almost always cold or campaign-shaped mail rather than replies.
Continuing while you investigate is how a two-week recovery becomes a two-month one, because every additional bad message resets the clock you are trying to run down.
Diagnose properly, in this order
Authentication. Check a real delivered message rather than the DNS. Alignment failures cause placement problems that look exactly like reputation damage and are fixed in an afternoon, per SPF, DKIM, and DMARC for an agent subdomain.
Which provider. Gmail and Microsoft disagree, and a problem at one is a different situation from a problem everywhere, per knowing where your agent's mail actually lands.
Complaints or bounces. Postmaster tools plus your own events. A complaint spike and a hard bounce spike have different causes and different fixes.
What changed. Volume, a new category, a prompt change that made messages more templated, a list from somewhere. Something did, and it is usually recent.
Blocklists. Check the major ones. Listing is less common than people assume and worth ruling out, and delisting is usually a form and a wait once the cause is fixed.
| Symptom | Likely cause | Fix |
|---|---|---|
| One provider only | Reputation at that provider | Volume down, engagement up, wait |
| All providers, sudden | Authentication or a listing | Check a real message, check lists |
| Rising hard bounces | Address source went bad | Fix sourcing, suppress, stop |
| Rising complaints | Unwanted mail | Stop that category entirely |
| Gradual decline | Volume grew faster than trust | Ramp back down and rebuild |
Repair or start again
The question everyone asks, and the answer is usually repair.
Repair the existing subdomain when the cause is identified and stopped, the domain has real history, and the damage is placement rather than outright rejection. Weeks of disciplined sending recovers it, and you keep the history you built.
Start a new subdomain when the damage is severe and old, when the cause was systemic rather than an incident, or when you need working delivery sooner than repair allows. It is not a reset button: the new subdomain has no reputation at all, so you are trading a damaged history for none, and warming from zero, per a warmup plan for a new agent sending domain.
What never works is rotating subdomains repeatedly to stay ahead of the damage. Providers see that pattern, it is the shape of a sender with something to hide, and the parent domain accumulates the reputation of all of them.
The recovery schedule
Slower than the original warmup, because you are overwriting a signal rather than establishing one.
- Week one: replies only. Nothing cold, nothing bulk-shaped. Mail people asked for, which they open and answer, which is the strongest positive signal available.
- Week two: same, slightly more. Watch per provider, not in aggregate.
- Week three: begin increasing, while complaints stay at zero and hard bounces near it. Hold the moment either moves.
- Week four onward: gradual growth, and only then reconsider whether the category that caused the damage should exist at all.
An inbound-first agent is unusually well placed here, since its normal traffic is exactly the engagement-heavy mail that rebuilds trust. That is a real advantage: a support agent doing its ordinary job is a reputation repair programme.
What does not help
- Rewriting the content. Content is the smallest lever and the first one everyone reaches for.
- Warmup services that exchange mail between seed accounts. Increasingly detected and discounted.
- Sending more to prove you are legitimate. The opposite of the required behaviour.
- Switching provider without changing behaviour. Domain reputation follows the domain, not the sender.
- Asking recipients to allowlist you. Fine as a supplement, useless as a strategy.
Preventing the second one
Recovery is a good moment to install what would have caught it: alerts on complaint rate above zero, on hard bounces rising, and on volume moving sharply, all routed to a person rather than a dashboard, per what to monitor in production.
And the structural fix, if the cause was outbound volume: keep it on a separate subdomain from the mail you need delivered, so the next incident cannot take your support replies with it, per deliverability for agent senders.
Questions
- How long does email reputation recovery take?
- Typically several weeks of disciplined sending, longer than the damage took. There is no way to accelerate it, and sending more makes it slower.
- Should I switch to a new subdomain?
- Usually not. Repair when the cause is stopped and the domain has real history. A new subdomain has no reputation at all, so you are trading damage for zero.
- Can I rotate subdomains to stay ahead of it?
- No. That pattern is recognisable, reads as evasion, and the parent domain accumulates the reputation of every subdomain you burn.
- What should I send during recovery?
- Replies to people who wrote to you. It is the strongest positive engagement signal, which is why an inbound-first agent doing its ordinary job repairs a domain as a side effect.
- Does changing provider help?
- No. Domain reputation follows the domain. Changing provider without changing behaviour changes nothing.
- What is the first thing to check?
- Authentication on a real delivered message. Alignment failures look exactly like reputation damage and are fixed in an afternoon.
Give your agent an address it can answer from.
Create an inbox