Key takeaways
- Hold AI-written outreach to the same honesty and consent standards as anything a person sends.
- Every fact in an AI draft needs checking, especially claims about the prospect.
- Personalise from public, professional information, never from anything that feels like surveillance.
- A named person stays accountable for each message, whoever or whatever drafted it.
What makes AI outreach unethical
AI outreach becomes unethical when it misleads, intrudes or ignores consent. The usual failures are invented details, false familiarity, volume that ignores relevance and data used in ways the person would not expect.
None of these are new. AI makes them cheaper to do at scale, which is why the rules need writing down before the tool is switched on.
- Claiming a referral, a meeting or a shared contact that never happened.
- Praising a post or project the prospect did not write.
- Faking a reply thread, such as a subject starting with Re: on a first email.
- Using personal details, such as family or health, that have nothing to do with the purchase.
- Carrying on after someone has asked you to stop.
Should you tell prospects that AI wrote the email
You should never pretend a person did work they did not do, and you should answer honestly if a prospect asks. Most teams treat AI like a template or a spell checker: the sender reviews the email, stands behind it and sends it under their own name.
The line is crossed when AI pretends to be a human in a live exchange, such as an automated agent that claims to be a named seller. If an assistant replies for you without review, say so in the message.
What data is fair to use for personalisation
Fair data is public, professional and relevant to the reason you are writing. An announced expansion, a job posting, a talk the prospect gave or a product they launched are all reasonable openers.
Ask one test question: would the prospect be comfortable seeing how you found this? A reference to their published article passes; a reference to their holiday photos does not. Keep only the data you need, and delete what you no longer use.
Which rules apply to AI sales emails
The same privacy and marketing rules apply whether a person or a model writes the email. In the UK and the EU that means data protection law, including a lawful basis for using contact data, plus the rules on electronic marketing; the US has its own rules for commercial email.
Across these regimes, the shared basics are a clear sender, an easy way to opt out and prompt handling of every opt-out. The details differ by country and by whether you write to a company or an individual, so agree your approach with whoever handles legal questions before you scale.
How do you keep a person accountable
Name an owner for every campaign and have them approve the copy and the list before anything goes out. If something goes wrong, one person should be able to explain why the message was sent.
While you are learning how a tool behaves, set it so outbound actions wait for approval, and keep a short record of what was reviewed and by whom.
How to set up ethical AI outreach, step by step
Agree the rules first, then set the tools up to follow them.
- Write a one-page outreach policy: what may be claimed, which data may be used and who approves sends.
- Check the lawful basis for contacting each segment, and record it with the list.
- Give the AI assistant verified company facts to work from, so it has no reason to invent claims.
- Review every draft for false familiarity, invented details and claims you cannot support.
- Include a clear sender and an easy opt-out, and stop contact as soon as someone asks.
- Check a sample of sent messages each month and update the policy with what you find.
Where Pebbles Ai fits
- CapabilityAI workspaceIn Manual mode, actions that reach the outside world, such as email sends, CRM writes and starting lead generation, wait for your approval; template send always asks, in every mode.
- FoundationCompany IntelAssistants write from Company Intel, the business knowledge and brand voice your team keeps, and your company information is never used to train an AI model.
Common questions
Is it legal to send cold emails written by AI?
Who wrote the email does not change the law; the same rules apply as for any marketing email. What matters is your lawful basis for contacting the person, a clear sender and an easy way to opt out.
Can AI personalisation feel creepy?
Yes, when it uses details the prospect would not expect a stranger to know or mention. Stick to professional, public information that explains why you are writing.
Who is responsible if an AI email contains a false claim?
The company that sent it, and in practice the person who approved it. That is why every AI-drafted message needs a named reviewer.
Should AI reply to prospects automatically?
Only for low-risk messages you have tested, and only if you are honest that the reply is automated. Replies that negotiate, promise or discuss price should come from a person.
About this page. Written by the Pebbles Ai team for people who run sales, marketing and growth. Last updated .
