ChatGPT prompts for LinkedIn posts that sound like you

The ChatGPT prompts that work for LinkedIn posts all do the same thing: they hand the model a specific fact from your work and a short set of voice rules, then ask for one post with a clear shape. A prompt that only says 'write a LinkedIn post about leadership' gets a post anyone could have written.

Below are ten prompts you can copy, each with what it is for. After them: how to give the model your facts and voice so you stop repeating yourself, and the three places the model will let you down. Replace everything in square brackets with your own material.

Set up the chat before the first prompt

ChatGPT has projects and memory, which let it carry context across chats. Put your voice rules and your reader in the project instructions once, instead of pasting them into every prompt. Keep them short: five or six plain rules and one line about who reads your posts. The first prompt below writes those rules for you.

Ten ChatGPT prompts for LinkedIn posts

• Voice rules from your own writing. 'Here are three LinkedIn posts I wrote and liked: [paste]. Describe my voice as five short rules I can reuse: sentence length, words I use, words I avoid, how I open, how I end. Plain words, no adjectives about tone.' Use it once, then paste the rules into your project instructions.
• Fact first draft. 'Here are three things that happened in my work this week: [fact one], [fact two], [fact three]. Pick the one a [reader, for example a founder of a small software company] would learn most from. Write a LinkedIn post of 120 to 200 words built on it. Put the fact in the first line. Short sentences, one idea each. No hashtags, no emoji, no question at the end.' Use it when you have material and no angle.
• Framework post. 'Turn this lesson into a framework post: [lesson]. Give it three to five numbered steps, each one a sentence a reader could act on tomorrow. Open with the situation where the framework helps, in one line. Under 220 words.' Use it for advice you give clients often.
• Receipt post. 'Write a LinkedIn post reporting this result: [what changed], from [number before] to [number after], over [period], because [what was done]. Use only these numbers. If the post needs a number I have not given you, write [NUMBER NEEDED] in its place. Do not estimate.' Use it for results, and keep the bracket rule in every version.
• Answer a client question. 'A client asked me this week: [question]. Write a LinkedIn post that answers it in under 150 words, for a smart reader outside my field. Start with the answer, then the reason, then one example from the details below: [details].' Use it for the questions you answer on calls anyway.
• Better openings. 'Here is a draft post: [paste]. Write five alternative first lines. Each must name something concrete from the draft: a number, a type of client, a decision. No questions, no cliffhangers, under 20 words each.' Use it when the post is right but the start is flat.
• Cut by a third. 'Cut this post by a third without losing any fact, name or number: [paste]. Remove warm-up sentences and any closing line that repeats the point. Keep my wording where you can.' Use it on every draft longer than you would read yourself.
• Ideas from your themes. 'My LinkedIn posts cover four themes: [theme], [theme], [theme], [theme]. Give me ten post ideas spread across them, each tied to something a person in my role would see first hand. Mark the ones that need a number from me.' Use it on the day you plan the week.
• Ideas from an article. 'Here is an article I wrote: [paste]. Find the three ideas in it that could stand alone as LinkedIn posts. For each, give a one-line summary and the sentence from the article it rests on. Do not write the posts yet.' Use it to get a week of posts out of one piece.
• Editor pass. 'Read this draft as a strict editor: [paste]. List every sentence that could appear in anyone's post, every number with no source, and every word I would not say out loud to a client. List the problems only, no rewrite.' Use it last, before anything goes out.

How to give ChatGPT your facts

Specific input produces specific posts far more reliably than any instruction about style. The model cannot know what happened in your week, so tell it.

1. Keep a running note during the week: questions clients asked, numbers you saw change, decisions you made and why.
2. Write each fact as one sentence with the detail in it. 'A client cut onboarding from nine steps to four' beats 'we improved onboarding'.
3. Give the source with every number: the dashboard, the invoice, the report.
4. Paste the facts at the top of the prompt, before the instruction.
5. Tell the model to use only those facts, and to leave a bracket where a fact is missing.

How to give it your voice

Voice rules work best as a short list of plain instructions with an example or two. 'Short sentences, one idea each. Name the client type. No exclamation marks. End on the practical point. No closing question.' Each time you fix the same thing in a draft twice, add a rule for it, so the list grows from your own corrections.

Where ChatGPT lets you down

ChatGPT is a capable writer. The problems below are about what happens around the writing, and they are the reasons most people who post with it stop after a few weeks.

• It forgets corrections between chats. Memory decides for itself what to keep, and a fix you made in Tuesday's chat may not reach Thursday's. Put lasting rules in the project instructions, and check them there, not in your head.
• It invents numbers. Ask for a result without supplying one and you may get a plausible figure nobody measured. The bracket rule in the receipt prompt is the defence, and it only works if you keep it in every prompt.
• It writes 'I' claims you never made. 'I spent years learning this the hard way' reads well and is a sentence you did not write about a thing that may not have happened. Read every first person line and delete any that did not come from your notes.
• It has no routine. Nothing reminds you on Monday, and the copy, the paste, the image and the timing are all still yours.

A last check before you post

• Every number has a source you could name if someone asked.
• The first line says something specific.
• No sentence claims a feeling or an experience you did not have.
• It stays well inside LinkedIn's 3,000 character limit and reads cleanly on a phone.

How GoodSocials handles this

GoodSocials works like these prompts with the gaps closed. A revise comment rewrites the card and updates one of three brand principles, voice, visual or format, which you can read in settings, so a correction stays made. When a post needs a number only you know, the card names it and the draft leaves a bracket, and a quality gate holds any post with an invented number before it can be scheduled.

Posts are written in the third person by rule, naming the company and the source, so nothing claims an experience you did not have. Approved posts publish at 09:00 in your timezone, one a day.