How to Write LinkedIn Posts With AI Without Sounding Like AI

Write the substance yourself: the story, the opinion, the number. Rough notes or a two-minute voice memo will do. Then use AI only to structure and tighten it, telling it to keep your sentences. Before posting, strip the tells: one-line paragraphs stacked like a staircase, "Let that sink in", lists of three, "I'm thrilled to announce", emoji bullets and a closing question fishing for comments.

The style isn't really what gives AI posts away. LinkedIn readers scroll past hundreds of posts a week, and the posts they skip are the ones with nothing specific in them: no number, no named situation, no opinion anyone could disagree with. A model can polish sentences, but it can't supply the week you had. Start from that and most of the "sounds like AI" problem disappears before you open the tool.

Follow me on Instagram@sagnikteaches

The tells that make a LinkedIn post read as AI

Here are the patterns readers notice, with the fix for each. Keep this table next to you for the first month.

Connect on LinkedInSagnik Bhattacharya

You'll find many of the same words in any list of AI clichés. A banned-phrase list for your brand turns this table into a standing instruction you paste into every prompt.

Subscribe on YouTube@codingliquids

Start with the substance: a two-minute voice memo

As a running example, take the chief executive of a small charity supporting young carers. It has eight staff, and she uses LinkedIn to reach funders, corporate partners and other charity leaders. She doesn't have time to write, but she has time to talk while walking between meetings.

After a trust turned down a bid, she recorded this memo (transcribed, lightly trimmed, illustrative):

"So we didn't get the [trust] grant. Sixty thousand over two years, and honestly we thought we had it. The feedback was really useful though. They said our outcomes weren't measurable. And they were right, we were counting sessions, not change. So we sat down as a team and we've started asking young carers three questions at the start and after twelve weeks. Stuff like, do you have someone to talk to, do you get time for yourself. Early days, but after the first group, eleven of fourteen said they had more time for themselves. We're reapplying in the spring. I think other small charities count activity because it's easy. I'd rather count the thing that matters even if it's messier."

Everything a good post needs is already there: a real setback, a specific number, what they changed, an early result, and an opinion other charity leaders might argue with. AI didn't need to add anything. It needed to shape it.

The tightening prompt that keeps your words

Below is a transcript of a voice memo. Turn it into a LinkedIn post.
Rules:
- Keep my sentences and phrases wherever possible. You are editing,
  not writing. Do not add facts, examples, quotes or statistics.
- Short paragraphs (2-4 sentences). No one-line-per-sentence format.
- No emoji, no hashtags in the body, no closing question.
- Don't use: thrilled, excited to share, journey, resilience, game,
  "let that sink in", "here's what I learned", lists of three adjectives.
- First two lines must make sense on their own (they show before
  "see more").
- 150-250 words.
Then list every change you made that wasn't just trimming.
[paste transcript]

The last instruction matters. Asking the model to list its non-trivial changes shows you exactly where it drifted from your words. An illustrative list returned with the draft:

Changes beyond trimming:
1. Opened with "We didn't get the grant." (moved from mid-memo)
2. Changed "Stuff like" to "for example"
3. Added "a humbling moment" after the rejection   <- remove
4. Combined the last two sentences into one

Change 3 is the model's instinct to add emotion you didn't express. Delete it. Changes 1, 2 and 4 are fine; that's what a good editor would do.

The statistic that wasn't in the memo

A more serious drift to watch for: numbers the model supplies to make a point sound weightier. In an earlier run with a looser prompt, the draft opened "With one in five young people now caring for a family member, measuring impact has never mattered more." The memo contained no such figure, and the chief executive couldn't say where it came from. Published, it would have been the sentence a funder checked first. The rule in the prompt ("Do not add facts, examples, quotes or statistics") exists for this, and the change list is how you catch it when the model ignores the rule. If you want a sector statistic in a post, find it yourself from a source you can name, and put the source in the post.

The first two lines do all the work

LinkedIn shows only the opening of a post before "see more": roughly 140 characters on a phone and a little over 200 on a desktop, depending on line breaks. Those two lines decide whether anyone reads the rest. Weak and strong openings for the same post:

• Weak: "I want to share some reflections on a recent experience that taught me a lot about leadership." (Says nothing; could be anyone.)
• Weak: "Rejection. 😔 It's never easy." (Staircase drama, no information.)
• Strong: "We didn't get a $60,000 grant we were sure of. The feedback was that our outcomes weren't measurable, and they were right."
• Strong: "For three years we counted sessions. A funder told us that proves activity, not change. So we've stopped."
• Strong: "Eleven of fourteen young carers told us they now get more time for themselves. Here's how we started measuring that."

The strong ones share a pattern: a fact, a tension and a promise that the rest of the post explains. Ask the AI for five openings in that pattern from your memo, then pick or rewrite one. Don't let it choose, because it tends to pick the most dramatic.

A full post, before and after

Here is the version you'd typically get by asking a model to "write a LinkedIn post about learning from a grant rejection", followed by the memo-first version the chief executive posted. Both are illustrative.

Topic-first (AI from scratch):

Rejection is never easy. 💔

But it's often our greatest teacher.

This week, our charity received news that a funding application wasn't successful.

At first, it stung. Then we got curious.

Here are 3 lessons I learned: 🔑 Feedback is a gift. 🔑 Measure what matters. 🔑 Resilience is a team sport.

What's the biggest lesson rejection has taught you? 👇

Memo-first (posted):

We didn't get a $60,000 grant we were sure of. The feedback was that our outcomes weren't measurable, and they were right.

We were counting sessions: how many young carers came, how often. That shows activity, not whether anything changed for them.

So we now ask three questions when a young carer joins and again after twelve weeks, for example "Do you have someone to talk to?" and "Do you get time for yourself?". After the first group, eleven of fourteen said they had more time for themselves. Early days, and a small group, but it's the first number we've had that measures what we exist for.

We're reapplying in the spring. I suspect plenty of small charities count activity because it's easy. I'd rather count the thing that matters, even if it's messier.

The first version could be posted by any organisation on earth; the second only by this one. It also does the charity's job: funders and partners reading it learn that the charity takes evidence seriously, which is the point of posting at all.

Five kinds of post a small organisation can keep up

A sustainable LinkedIn presence doesn't need daily posts. It needs a few repeatable types you can draw from each week:

1. A lesson with evidence. Something that went wrong or right, what changed, one number. (The grant post above.)
2. Behind a number. Take one figure from your annual report and explain it. "We supported 140 young carers last year. Here's what one week of that looks like for our two support workers."
3. A specific thank-you. Not "huge thanks to our amazing partners", but what the partner did and what it made possible, with their permission to tag them.
4. A view on your sector. A clear opinion on something other leaders face. These get the most discussion and need the most care.
5. A vacancy or volunteer call. The role, what the day looks like, who'd enjoy it. Post from the Page and share from your profile.

Here is type 3 done badly and well (illustrative):

Before: "Huge thanks to our amazing corporate partners for their incredible support this year! 🙏 We couldn't do it without you! #grateful #charity"

After: "A local accountancy firm gave us 40 volunteer hours this year. Two of their team rebuilt our budget spreadsheet, and that's the reason our last bid had a cost breakdown a funder could actually follow. Thank you, [firm]."

The after version tells other businesses what volunteering with the charity looks like, which is the most useful thing a thank-you can do.

Each type suits the memo-first method. For the sector view, record your opinion first, before asking the AI anything, so its ideas don't replace yours. Then use the model as a sparring partner rather than a writer: "Here's my view in a paragraph. Give me the three strongest objections a thoughtful charity leader would raise." If one objection is fair, address it in a sentence. The post becomes harder to dismiss, and it's still your argument. For the thank-you, include what the partner actually paid for or gave, because vague gratitude reads as template.

Where LinkedIn's own AI tools fit, and don't

LinkedIn has experimented with built-in AI writing help for posts, but its help page for the AI-powered post-writing tool currently says the tool "is not available at this time", with new tooling promised. In practice, that means drafting in ChatGPT, Claude or Gemini and pasting the result into LinkedIn. If you use an assistant regularly, save your voice notes, banned phrases and a few of your best posts in a Project so each draft starts from your voice rather than the model's default.

What LinkedIn does offer natively is scheduling: you can schedule a post from an hour ahead up to three months in advance, and Page admins can schedule posts for the Page. Scheduled posts can be rescheduled or deleted but not edited, so do the final read before you schedule, not after.

Comments on other people's posts are part of it

For a small organisation, thoughtful comments on other leaders' and funders' posts often do more than your own posts, because they reach an audience you haven't built yet. They're also where AI help is most visible. "Great insights! Thanks for sharing 🙌" and "This really resonates. Well said!" are the comments everyone recognises as automated or lazy.

Write comments yourself, in two sentences: one specific reaction and one thing from your own experience. For example, on a funder's post about outcome measurement: "The twelve-week check-in has been the easiest change for us, because young carers answer three questions in two minutes. Would be interested to hear whether other grantees found a shorter version works." If you use AI at all here, use it only to check the tone of a comment on a sensitive post, never to generate comments in bulk.

A 30-minute weekly routine, from voice memo to post

1. During the week (2 minutes each): record a voice memo whenever something happens worth telling. Most weeks you'll have two or three.
2. Once a week (15 minutes): pick the best memo, transcribe it (your phone's recorder or meeting tool can), and run the tightening prompt.
3. Edit (10 minutes): check the change list, delete anything the model added, rewrite the opening if it's weak, and read the whole post aloud.
4. Schedule (3 minutes): for a time your audience is at work; for most charity and business audiences, weekday mornings are a reasonable starting point to test.

Keep the memos you didn't use. A memo that wasn't the best this week is often the best next week, once a second thing happens that connects to it.

Checking a post doesn't sound like AI before you publish

Three quick tests, in order:

• The swap test. Replace your organisation's name with a competitor's. If the post still works, it's too generic; add a number, a name (with permission) or a detail from your week.
• The read-aloud test. Read it out as if to a colleague. Any sentence you'd be embarrassed to say out loud goes.
• The colleague test. Send it to one person who knows how you talk, with one question: "Does this sound like me?" If they hesitate, ask which line.

An illustrative example of the swap test catching something: an early draft of the vacancy post said "Join a passionate team making a real difference in young people's lives." Swap in any charity's name and it still works, so it failed. The replacement, "Our support workers spend most afternoons in school car parks and kitchens, not offices," could only be this charity. For the wider reasons AI copy drifts generic, see why AI marketing copy sounds generic; for getting the organisation's voice written down once, a brand voice guide AI can follow is the place to start.

If your organisation's other channels are also AI-assisted, the charity-specific advice in charity appeals and social posts that keep your voice applies to LinkedIn too. And if you're wondering whether to mention AI at all, whether to tell customers you use AI in marketing sets out when it's worth saying.

LinkedIn and AI: questions people ask next

Does it matter if people can tell AI helped with a post?

Readers don't object to help; they object to posts with nothing in them. A post built from your own story, numbers and opinion, tidied by AI, reads as yours. A post generated from a topic reads as generic whoever typed it. If someone asks, say honestly that you draft from voice notes and use AI to tighten. That's a normal way to work.

Should I post as myself or from the organisation's Page?

Both, for different jobs. Personal posts from a founder or chief executive usually get more conversation, because people respond to people. The Page is the right place for announcements, vacancies and news others can share. A workable split is one personal post a week with a view or a lesson, and one or two Page posts with facts and updates.

How long should a LinkedIn post be?

LinkedIn allows up to 3,000 characters, but most good posts from small organisations use far fewer, often 600 to 1,300. Long enough to tell one story or make one point with evidence; short enough that nobody needs to scroll twice. What matters most is the opening, because only the first two or three lines show before 'see more'.

Can I reuse my Instagram captions on LinkedIn?

Reuse the substance, not the text. An Instagram caption about an event can become a LinkedIn post about what the event taught you, what it cost, or what it achieved. LinkedIn readers want the reasoning and numbers behind the picture, so give the AI the caption plus two or three extra facts and ask for a version that explains why it matters.

Further reads

• How to Write Instagram Captions With AI That Sound Like You — The same voice-file method adapted for Instagram.
• How to Turn One Customer Story Into a Month of Marketing With AI — Turn one story into several LinkedIn posts and more.
• How to Write Case Studies From Customer Interviews Using AI — Case studies give LinkedIn posts their strongest evidence.
• How to Write AI Prompts That Sound Like Your Business — Prompts that keep every channel sounding like you.
• How to Find B2B Prospects With AI Research Tools — Find the people you want your posts to reach.
• How to Write a Press Release With AI That Journalists Will Read — When a post's news deserves a press release too.
• AI for Recruitment Business Development: Winning New Clients — A weekly business-development routine for agencies: hiring signals, five-minute account briefs, outreach that names the vacancy, and consented spec profiles.
• Should a Small Business Use AI to Write Its Social Media Posts? — A decision grid for which social posts to hand to AI, three businesses with three different answers, and a four-week trial to test it properly.
• Social Media Pre-Publish Checklist for AI-Drafted Posts — A grouped pre-publish checklist for AI-drafted social posts, with why and how to check each item, platform limits, AI labels and a worked example.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: LinkedIn Help pages on post scheduling and on its AI-powered post-writing tool (listed as not available, September 2026); LinkedIn post length limit; general writing practice.