AI can write a newsletter in seconds. The problem is that it often sounds like everyone else’s newsletter. Content that sounds like you takes more than a clever prompt: it takes your own examples, a few clear rules, and a person who approves every piece.
Here is why AI content so often sounds generic, what your tone of voice is made of, and how an agent learns it.
Why AI content sounds generic
Out of the box, an AI writes like an average of everything it has read: polite, smooth and easy to forget.
the same phrases everyone uses, like “in today’s fast-paced world”
big promises without proof
no stories, examples or numbers of your own
the wrong level of formality for your customers
in Danish, German, French or Dutch, sentences that read like a translation
Asking the AI to “sound friendly” doesn’t fix any of that. It needs to know what you sound like.
What your tone of voice is made of
Your tone of voice is a handful of choices you make every time you write.
Words: the terms you use and the ones you avoid, like “customers” or “clients”.
Rhythm: short and direct, or calm and explanatory.
Formality: first names or titles; “du” or “Sie” in German, “tu” or “vous” in French.
What you never say: claims you won’t make, and words that feel wrong for your brand.
Your proof: the stories, examples and results only you have.
How an agent learns your voice
An agent learns your voice from your own material, a few written rules and your corrections.
Your best examples. We collect texts you are proud of: emails, posts, newsletters and your website.
A short style guide. The words you use and avoid, how formal you are, and what you never claim.
Your facts. Your services, prices, policies and stories, so the agent writes about your company and not a generic one.
Your corrections. The edits you make in the first weeks become rules for the next drafts.
After a few rounds, the drafts need fewer and smaller edits.
Why you approve first
Nothing goes out without your approval, because you are responsible for what is published in your name.
Facts: AI can get details wrong, so someone who knows the business checks them.
Voice: reading every piece keeps the tone on track, and you catch drift early.
The law: since 2 August 2026, the EU AI Act (Article 50) requires AI-generated text published to inform the public on matters of public interest to be disclosed, unless a person has genuinely reviewed it and someone takes editorial responsibility for it.
What Google says about AI content
Google allows AI-assisted content, as long as it is accurate, useful and made for people.
Its guidance on generative AI content warns against using AI to generate many pages without adding value, which may violate its spam policy. It also suggests that telling readers how a piece was made can give them useful context. So the test is the same as for any content: would your customer be glad they read it?
Questions and answers
Can AI write in Danish, German, French or Dutch?
Yes, and well, if it gets examples in that language. A native speaker should still read it before it goes out, and a translation should be adapted to the language, not done word for word.
Do I have to say a text was written with AI?
For ordinary marketing content, no general label is required. Google says explaining how content was made can help readers. For texts that inform the public on matters of public interest, the AI Act requires a label unless a person has reviewed the text and takes editorial responsibility for it. AI-made images, audio or video that could pass as real (deep fakes) must be labelled.
How much of my time does it take?
Setting up your voice takes a little time at the start. After that, reviewing a draft takes far less time than writing it yourself.
What does Acastia do here?
Our marketing agents write articles, posts and newsletters in your tone of voice, and you approve every piece before it goes out. See marketing, content and SEO.
