How-To Guide

How to Use AI for LinkedIn Without Sounding Like a Bot: A Guide for Small Businesses and Consultants

A 90-minute weekly workflow, ready-to-use prompts and the rules that keep your LinkedIn posts sounding like you — not like everyone else’s ChatGPT.

B Biztrategy Published 17 November 2026 · 8 min read
A cozy workspace setup featuring a laptop, coffee, and a lamp on a checkered tablecloth.

You open LinkedIn, scroll for ten seconds and spot it straight away: the post that starts with a one-line hook, breaks into single-sentence paragraphs, lists “5 lessons I learned” and closes with “Agree? 👇”. Half your feed now reads like it came out of the same prompt — because it did.

That is the problem for consultants, freelancers and small business owners. LinkedIn is often the single best source of B2B leads for a small firm, and AI makes it far easier to post consistently. But if your posts sound like everyone else’s, consistency buys you nothing. Buyers skim past generic content, and the people who matter most — prospects who already half-know you — notice when the voice is not yours.

This guide shows how to use AI for the parts of LinkedIn it is good at (ideas, structure, editing, repurposing, research before outreach) while keeping the parts that win work (your opinions, your stories, your numbers) firmly human. It includes a weekly workflow that takes about 90 minutes and the prompts to run it.

What AI should — and shouldn’t — do on LinkedIn

The owners we speak to tend to fall into two camps. One group refuses to touch AI for LinkedIn because “it all sounds the same”. The other lets ChatGPT write every post and wonders why nothing converts. Both are leaving value on the table. The useful split looks like this:

  • Good jobs for AI: turning a voice note or meeting transcript into a draft, suggesting angles on a topic, tightening a rambling post, rewriting a hook five ways, repurposing a blog post or case study into three posts, summarising a prospect’s recent activity before you reach out, and spotting patterns in which of your posts performed.
  • Jobs to keep human: the opinion, the story, the client result, the numbers, the final read-through, and every one-to-one message and comment reply. These are what make someone think “this person actually knows my problem”.
  • Jobs to avoid entirely: automated connection requests, auto-DMs and bots that like or comment on your behalf. LinkedIn’s terms prohibit most third-party automation of this kind, accounts do get restricted, and prospects can tell.

A simple rule: AI can shape the words, but the substance must come from you. If a post contains no detail that only you could know, it is not ready.

Start with raw material, not a blank prompt

The single biggest reason AI LinkedIn posts sound generic is the input. “Write a LinkedIn post about AI for accountants” gives the model nothing except the average of the internet, so that is what you get back.

Instead, feed it something only you have. Good sources of raw material for a small firm:

  • Voice notes. Record two or three minutes on your phone after a client call: what surprised you, what the client was worried about, what you told them. Transcribe it (most phones and AI apps do this for free) and use the transcript as the input.
  • Client questions. Every question you answered by email this week is a post. Keep a running list.
  • Proposals and reports. The “why we recommend this” section of a proposal is usually a ready-made opinion piece, once you strip out the client’s details.
  • Your own numbers. “We cut a client’s quoting time from three days to four hours” beats any statistic the model can invent.
  • Disagreements. Where do you think common advice in your field is wrong? That is your most shareable content.

A useful habit: keep a single note on your phone called “LinkedIn raw” and drop one line into it every time a client says something interesting. By Friday you will have more material than you can use.

A 90-minute weekly LinkedIn workflow

For most consultants and small business owners, three good posts a week plus 15 minutes a day of genuine engagement is plenty. Here is how to batch the writing in one sitting.

Step 1: Pick three ideas (15 minutes)

Open your “LinkedIn raw” note and paste it into Claude or ChatGPT with this prompt:

Prompt: “Below are my notes from client conversations this week. I am a [your role] who helps [your client type] with [your outcome]. Suggest six LinkedIn post angles drawn ONLY from these notes. For each, give the core point in one sentence and say whether it is a story, an opinion, a how-to or a lesson. Do not invent facts, numbers or clients. Notes: [paste]”

Choose three that you would be happy to defend in a sales call. Aim for a mix: one story, one opinion, one practical how-to.

Step 2: Talk, then draft (30 minutes)

For each idea, speak for two minutes into a voice recorder as if explaining it to a client. Paste the transcript in with:

Prompt: “Turn this transcript into a LinkedIn post of 150–250 words. Keep my phrasing and examples wherever possible — edit for clarity, do not rewrite my voice. British English. No emojis, no hashtags in the body, no ‘Here’s the thing’, no rhetorical question to open, no ‘Agree?’ at the end. Short paragraphs, but not one sentence per line. Transcript: [paste]”

The banned-phrases list matters. Every model has verbal tics, and LinkedIn readers have learned to spot them. Add to your list every time you notice one you dislike. Our guide on making AI writing sound like your brand explains how to turn this list into a reusable style guide.

Step 3: Sharpen the hook (15 minutes)

The first two lines decide whether anyone clicks “see more”. Ask for options, then pick or blend:

Prompt: “Give me five alternative opening lines for this post. Each must be specific and concrete — a number, a client situation or a clear opinion. No clickbait, no questions, no ‘I’ve been thinking about’. Post: [paste]”

Step 4: Human edit and fact-check (20 minutes)

Read each post aloud. Cut anything you would not say to a client’s face. Check every number against its source. Remove any client detail they have not agreed to share — anonymise by sector and size (“a 12-person logistics firm in Valencia”) rather than naming names. Then schedule the posts using LinkedIn’s built-in scheduler.

Step 5: Engage manually (10 minutes a day, not included above)

Reply to every comment on your posts yourself, and leave two or three thoughtful comments on posts by prospects and peers each day. This is where relationships actually form. AI can help you understand a long post quickly, but write the comment yourself.

Repurpose what you already have

If you already write blog posts, newsletters or case studies, you are sitting on months of LinkedIn content. One 1,500-word article typically yields three to five posts: the main argument, one or two practical tips, a client example and a contrarian point.

Prompt: “Here is an article I wrote. Extract four standalone LinkedIn posts from it, each making one point and each 120–200 words. Use only content and examples from the article. Write each one so it makes sense to someone who has not read the article. Article: [paste]”

Client case studies are especially strong on LinkedIn because they show results rather than claim them. If you do not have any yet, our guide on writing client case studies with AI walks through the interview and approval process. For a wider system covering newsletters, video and email, see how to use AI for content repurposing.

Using AI for outreach without spamming

Most people receive LinkedIn messages that are obviously templated: “I came across your profile and was impressed by your journey”. AI has made these cheaper to send at scale, which is exactly why they no longer work.

Where AI genuinely helps with outreach is preparation, not sending. Before contacting a prospect, copy their recent posts, their company’s “About” page and any recent news into a prompt:

Prompt: “I help [client type] with [outcome]. Based on the information below about this person and their company, list three specific challenges they might be facing that relate to my work, and one thing they posted recently that I could reasonably comment on. Flag anything you are guessing. Information: [paste]”

Then write a short, specific message yourself — three or four sentences, one clear reason for reaching out, no pitch in the first message. A realistic target for a small firm is five to ten personalised messages a week. That sounds modest, but a 20–30% reply rate on ten good messages beats a 1% rate on 300 automated ones, and it does not put your account at risk.

A note on data: pasting publicly available profile information into an AI tool for your own research is low-risk, but do not paste private messages or contact lists into free consumer tools, and check your provider’s data settings. If you are building a fuller prospecting system, our guide on using AI for lead generation covers the GDPR side in more detail.

Common mistakes to avoid

  • Posting the first draft. If you did not change at least a third of it, it probably still sounds like the model.
  • Letting AI invent stories. “A client told me last week…” must be true. Fabricated anecdotes are easy to spot and destroy trust instantly.
  • Chasing the format instead of the point. Hooks, line breaks and lists are fine, but a post with no clear point is still noise, however well formatted.
  • Using AI-generated comments. Generic “Great insights, thanks for sharing!” replies, human or machine, do nothing. One specific comment is worth twenty of them.
  • Measuring the wrong thing. Likes from peers are pleasant; enquiries from buyers pay the bills. Track profile views from your target sector, connection requests from prospects and conversations started. Once a month, paste your post analytics into AI and ask which topics drove profile visits rather than reactions.
  • Going quiet after a busy month. AI makes consistency cheap. Use the batching workflow to keep a two-week buffer of scheduled posts, so client crunches do not wipe out your visibility.

The bottom line

AI will not make you a LinkedIn thought leader, and it should not try to. What it does well is remove the friction: turning a two-minute voice note into a clean draft, giving you five better hooks, extracting four posts from an article you already wrote, and preparing you for a conversation with a prospect. Keep the opinions, stories and numbers yours, edit every post before it goes out, and do the engaging yourself. Ninety minutes a week, done that way, is enough to become the person your ideal clients think of first.

Where does your business stand on AI?

Take the free 3-minute AI Readiness Quiz and get a personalised score with your next steps.

Take the Free Quiz →