AI cold email in 2026: what actually gets replies (and what gets you blocked)
Short answer: the email that gets a reply is short, specific to one company, and sent from a warm, separate domain to a verified address. AI helps before the message (research, list cleaning) and after it (reading and routing replies). Letting AI write the message itself is how domains get blocked.
We run outbound for our own business and as a managed service, so we see both sides: what gets a reply and what gets a domain quietly throttled. The gap between the two has widened in the last year, and it is mostly about AI being used in the wrong place.
Where AI belongs
- Before the message. Reading a company's site to find one real fact — what they make, who they serve, what changed recently. That fact is the only personalisation that works, and it should be a real merge field, not a guessed one.
- Cleaning the list. Verifying addresses and removing catch-alls and role inboxes before anything is sent. Our own verifier only marks an address valid or invalid when it can prove it; a blocked lookup is never treated as invalid.
- After the reply. Reading responses and sorting them: interested, not now, wrong person, unsubscribe. A person only touches the interested ones.
Where it does not
Bulk-writing the message. Generated cold email has a texture that people recognise in a sentence: the brand-positioning opener (“At X we…”), the “happy to”, the long dash. We keep a written list of these tells and remove them. The message is in first person, plain language, one ask, and the signature is a first name with nothing under it.
The infrastructure most people skip
Outreach carries reputational risk, so it must be isolated. We learned this expensively: a domain that carried outreach and hosted infrastructure was suspended by its registry, and unrelated systems went down with it. Outreach domains are now always separate from anything that matters, and inboxes are warmed before they send a single real message. This is not optional; it is the difference between a campaign and a blocklist.
What the service includes
Two dedicated sending domains and six inboxes, a three-step sequence written for your offer, address verification, daily monitoring and a monthly report, for €300 a month with monthly cancellation. The public target is about 1,000 messages to 15 introduction meetings a month — a target, not a promise; list quality and how narrow the offer is decide the real number.
If you want to see how the sequence reads before you commit, book a 20-minute scoping call and bring the one thing you sell best. We will write the first email with you on the call.
Frequently asked questions
Does AI-written cold email work?
Fully AI-written bulk email is the fastest way to get a domain blocked: the text is generic, the patterns are recognisable, and reply rates collapse. AI works when it is used for research, list cleaning and reply routing, while the message itself stays short, specific and written in one clear voice.
Why do you send from separate domains?
Because outreach carries reputational risk. If a sending domain is flagged, it must never take your main domain and your infrastructure down with it. We learned this the hard way: a registry suspension on a domain that also hosted infrastructure took unrelated systems offline. Outreach domains are now always separate.
How many emails does it take to get a meeting?
Our public target for the Lead Generation service is roughly 1,000 messages to 15 introduction meetings a month. It is a target, not a guarantee; the number depends on list quality and how narrow the offer is.
What does managed lead generation cost?
€300 per month: two dedicated domains and six inboxes, a three-step sequence, daily monitoring and a monthly report. Monthly cancellation.
Do you verify the email addresses?
Yes, before sending. A list with dead addresses damages the sender reputation of every inbox in the campaign, so unverified addresses are not sent to.
— Written by the Bulsu Labs team · Talk to us