Question
How do I write LinkedIn posts in my own voice with AI?
Give the AI two or three paragraphs you wrote yourself, a few opinions you actually hold and a list of true facts about you, then edit every draft before it posts. Your paragraphs set the rhythm. The opinions give each post a point to argue, and the facts stop it inventing numbers. The edit is what makes it yours.
Why does AI writing sound generic in the first place?
A model with no samples of you writes the average of what it has read, and the average has no opinions.
Describing your tone with adjectives (“friendly but direct”) gives it very little to copy. Your own sentences give it far more: how long they run, the words you reach for, how you open, what you never say.
Then there’s the stance. Without one, a draft argues whatever sounds reasonable to everyone, which is why so many AI posts could have been written by anyone.
And there are the invented specifics. When a model has no figure for a claim, it can make one up, and a reader who knows you will catch it.
What should I give the AI before it writes anything?
Your own writing, your positions and facts anyone could check, plus a short list of words you’d never use.
| What to give | How much | What it fixes |
|---|---|---|
| Paragraphs you wrote | 2 or 3, your best LinkedIn posts work well | Rhythm and word choice |
| Positions | One sentence each, things you hold or reject | Drafts that sit on the fence |
| Proof facts | One per line, dated when the number changes over time | Invented figures |
| Content pillars | 2 or 3 subjects you want to be known for | Posts that wander off topic |
| Words to avoid | The ones that make you wince | Phrases that don’t sound like you |
A proof fact should be something a stranger could verify, like a founding year or a count you’d stand behind in public. Put a date on any fact that grows, so last spring’s number doesn’t go out as if it were new.
How do I stop the AI from making things up?
Give it the real numbers, and use a tool that marks a gap instead of filling it.
- Keep your facts in one place the AI reads every time, so you don’t paste them into each prompt.
- Ask for a visible placeholder wherever a figure is missing, and fix it before you post.
- Check every quote against where it came from. A quote you can’t find is a quote you shouldn’t publish.
- Watch the first line. A hook that opens on a statistic you never gave the AI is the most common place an invented number hides.
How much should I edit an AI draft?
Enough that you’d say every sentence out loud to a buyer. We budget about five minutes a post, and the weekly routine treats that edit as the step worth protecting.
- Cut the opening paragraph if it warms up instead of starting.
- Add one thing only you know: what a customer said on Tuesday, the number from your week, the mistake you made.
- Delete any sentence you wouldn’t say in person.
- Read the first line on its own. LinkedIn shows the start of a post and hides the rest behind “see more”, so that line carries the post.
- Drop a closing question that only exists to fish for comments.
A draft you edited for five minutes sounds like you. A draft you approved in ten seconds sounds like the tool, and people who know you will notice.
Can I just use ChatGPT for this?
Yes, for the odd post. The cost shows up later, when you’re pasting your samples, facts and opinions into a fresh chat every week and checking each number by hand.
If you write once a quarter, a general assistant is the cheaper choice. We say the same on our Adamiro vs. ChatGPT page. If you write every week, the setup is the part worth keeping somewhere permanent.
How Adamiro does it
Adamiro keeps your voice as a saved profile in Voice and uses it on every draft.
- Teach it. Open Voice and go to Your Voice. Press Teach it, find Work it out from writing you already have, paste a few paragraphs and press Read my writing. Adamiro proposes a personality and tone for you to check, and nothing is stored until you press Save. It also keeps the paragraphs you pasted, and every later draft in that language is written with up to three of your own passages in view.
- Add your facts and subjects. The same screen holds Content pillars and Proof facts. Voice states your proof facts word for word and leans each draft toward a pillar when the topic fits.
- Write down what you think. Positions you type are used straight away. Adamiro also suggests positions from your Mentor conversations and the sentences you cut from drafts, each shown with the words that led to it, and none reaches a draft until you accept it.
- Write the post. On the Ideas tab, press Write as… on an idea, choose LinkedIn Post, and set Write as to You or The company. Voice writes in any of 32 languages.
- Check the gaps. Where Voice has no figure for a claim, it writes a marked fact to verify and lists it, and Reach won’t approve the post until you’ve replaced it. Words on your Never use these list are removed after the draft is written.
- Try another opening. Other hooks offers three more first lines and drops any that carry a number you never supplied.
- Edit and approve. Press Approve for Reach. Reach publishes to your LinkedIn profile or Company Page only after a person approves the post.
Your edits feed back in. Adamiro notices the words you keep removing and proposes changes to your voice, and your voice changes only when you press Use it. Every rule Voice follows is listed on the Voice features page. Drafts are paid from a task balance, and the pricing page has the numbers.
If a whole team is posting, our playbook on why your team should write their own LinkedIn posts covers company voice versus each person’s own.
When Adamiro is the wrong tool
- You want posts to go out without review. Every post in Adamiro waits for a person to approve it. There’s no setting that turns that off.
- You want engagement on autopilot. Adamiro doesn’t like or comment on your behalf. Voice can write a comment for you to paste under someone’s post, and you post it yourself. If automated LinkedIn engagement and a following-growth toolkit are what you’re after, Taplio does more on LinkedIn itself than we do.
- You want X posted for you. Voice writes X threads, and you paste them in by hand. Adamiro doesn’t post to X.
- You post a few times a year. A general chat assistant will do the job for less.
Questions people ask
Will people be able to tell AI wrote my post?
If you post the first draft untouched, often yes. The tells are generic openings, invented numbers and endings that beg for comments. Adding one detail only you know, and cutting what you wouldn’t say aloud, removes most of them.
How much of my own writing does the AI need?
Two or three paragraphs are enough to start. Your best LinkedIn posts are a good choice.
Can each person on my team have their own voice?
Yes. On a Team account each Co-Editor keeps their own voice, pillars and proof facts. The company voice is a separate profile that the owner and group admins edit, and a post goes out through the account that matches the voice that wrote it.
Does it work in languages other than English?
Voice writes in 32 languages. Word lists are kept per language, so a rule you set for Swedish never touches an English draft.
Is my writing used to train AI models?
Nothing you write or Adamiro generates trains a model without your written consent, and consent is off until you give it.
Does an approved post carry an AI label?
Published text can carry an AI disclosure line, and you control it per account.
Last updated: 2026-10-18