Yes. No mailbox provider bans AI-written email and no filter detects it as such. The real constraints are that it is good at some tasks and poor at others, that recipients have learned to recognise the style, and that it makes no difference whatever to whether your message arrives.
Published by emailcampaign.ai. Our product includes AI drafting, so the limits below are worth reading with that in mind.
What it is good and bad at
| Task | Suitability | Why |
|---|---|---|
| Summarising a company site into relevant facts | Strong | Reading and compression are the core capability |
| Spotting a trigger in a job posting or announcement | Strong | Pattern recognition across unstructured text |
| Tightening something you wrote | Good | Editing outperforms generating |
| Drafting several angles to choose between | Good | Useful raw material with you as the filter |
| Sorting and triaging replies | Good | Classification is a well suited task |
| Writing the opening line unsupervised | Weak | It cannot judge whether an observation matters |
| Writing a whole cold email | Poor | Fluent, generic, and recognisable |
The pattern is that it is strong wherever the work is reading and weak wherever the work is judgement. Deciding which fact about a company implies something worth writing about requires knowing what your offer does and who feels the problem, which is not information the model has.
The failure mode
The characteristic bad output is accurate and pointless. "I noticed you recently expanded into the Nordics" is true, easily retrieved, and tells the reader nothing about why you wrote. It signals that something automated read their website, which is the opposite of what personalisation is meant to convey.
Useful personalisation connects an observation to a consequence for that person. The fact is the easy half. The inference is the half that requires you.
The tells recipients now recognise
People receive several AI-written emails a week and have learned the pattern. The markers are consistent:
- Em dashes, which appear far more often in generated text than in ordinary business writing.
- Openers such as "I hope this email finds you well" and "I noticed that".
- Uniform sentence length and rhythm through a whole message.
- A compliment that would fit any company in the sector.
- Lists that always have exactly three items.
One of these is survivable. Three in a short message reads as automated, and the reply rate reflects it.
Verify anything concrete
Never let a model produce a number, a customer name, or a claim about the recipient’s business without checking it. A confidently wrong detail about their own company is worse than no personalisation at all, because it demonstrates that nobody looked.
This applies to claims about your own product too. Fluent text is persuasive to the reader and to the person who generated it, which is exactly why unchecked specifics survive review.
It does not affect deliverability
Worth stating plainly, because the two get conflated constantly. Whether your email reaches the inbox is decided by the sending domain’s history, whether SPF, DKIM and DMARC actually resolve rather than merely exist, the hard bounce rate, the complaint rate against Google’s published 0.1 percent threshold, and how many messages each mailbox sends per day.
Authorship appears nowhere in that list. A perfectly written message from a cold, unauthenticated domain still lands in spam, and a plainly written one from a warmed, authenticated domain does not. If placement is the problem, check authentication first; it takes about thirty seconds and frequently finds a duplicate SPF record or an include chain past ten lookups.
A workflow that holds up
- Model reads the company site, recent announcements and any job postings.
- It returns two or three candidate facts relevant to your offer.
- You pick the one with a real implication, or skip the prospect entirely.
- You write the opening line. The model may tighten it.
- You verify every concrete claim before sending.
Step three carries the most weight. Being willing to drop a prospect because nothing relevant turned up is what keeps a list narrow, and list quality moves reply rate far more than copy does.
Questions, answered straight
- Can I use AI to write emails?
- Yes. No mailbox provider or email platform prohibits AI-written content, and filters do not detect or penalise it as such. What they score is sending behaviour: domain history, authentication, bounce rate and complaint rate, none of which change based on who wrote the text.
- Do I have to disclose that an email was written by AI?
- There is no general requirement to disclose it in ordinary business correspondence. The obligations that do apply to commercial email are about identifying the sender, not the drafting method. Some regulated industries have their own rules worth checking.
- Can recipients tell an email was written by AI?
- Increasingly, yes, and it is stylistic rather than technical. The recognisable markers are em dashes, openers like 'I hope this email finds you well' and 'I noticed that', uniform sentence rhythm, and compliments that would apply to any company.
- Does AI-written email hurt deliverability?
- No. Placement is decided by the sending domain's history, whether authentication resolves, hard bounce rate, complaint rate and volume per mailbox. Content is the smallest factor and authorship is not a factor at all.
- What should AI never write in an email?
- Anything factual it could invent: numbers, customer names, claims about the recipient's business, or specifics about your own product. Verify every concrete detail, because a confidently wrong fact about their company ends the conversation.
Published 22 September 2026. Updated 22 September 2026. Written by the team that runs the infrastructure; numbers come from the platform's own provisioning and sending, and from the providers' published documentation at the time of writing.