1d:I[58553,["2550","static/chunks/96cb50b6-60e4ebcfc9e986f2.js","8754","static/chunks/eed5e382-ee92683b99edf94b.js","2534","static/chunks/2534-d4488ec44c364215.js","7101","static/chunks/7101-f413d89e9cba3b63.js","1310","static/chunks/1310-e2a289329247ece1.js","6790","static/chunks/6790-8ccbcaf79fba421f.js","2642","static/chunks/2642-abf34e6aa57e0523.js","9230","static/chunks/9230-abec6248ed19d865.js","1566","static/chunks/1566-140144b37dd19b78.js","3244","static/chunks/app/(app)/blog/%5Bslug%5D/page-32be5b62ae91bd42.js"],"Image"] 1e:T1a2c,

You generate a LinkedIn post in ChatGPT and it is terrible. You read it and think: this is not me. This is fluff. This feels like a model.

The model is doing what you asked. You asked for a viral post and gave it no week, no scene, no belief. So it wrote the average of every viral post it has seen.

A better prompt helps. A better prompt is not the whole fix. The fix is feeding it something only you could have typed — then looking at the result as a LinkedIn post with a body, not as a chat bubble.

Before: the post ChatGPT writes from nothing

Prompt people actually use: `Write me a viral LinkedIn post about consistency.`

Output you actually get (this is the pattern, not a ranking claim):

> In today’s fast-paced world, consistency is the key to success. > > Here are 3 tips to grow on LinkedIn: > > 1. Post regularly > 2. Engage with your network > 3. Be authentic and add value > > The best time to start is now. > > What’s your #1 tip? 👇

That post has no person in it. No day. No cost. The numbered tips are slogans. The question at the end is a costume CTA. Readers have seen this shape a hundred times; they scroll.

If your drafts look like this, stop asking for viral. Start from a note.

After: the same topic, with a week in it

Same job (consistency). Different input: a real Tuesday.

> I deleted a draft this morning because the first line was “In today’s fast-paced world.” > > That sentence is how you tell two thousand people you have nothing to say. > > I had sat down to write about consistency. ChatGPT gave me three tips I already ignore: post regularly, engage, be authentic. I could have posted it in four minutes. I would have been one more person performing a system I do not run. > > What I actually did last week: three drafts on Thursday, two parked, one shipped Friday. Twenty minutes in comments on posts I would read if they were not mine. No daily streak. No 7 a.m. ritual. > > Consistency, for me, is a batch I can keep when the product is on fire. Not a quote about showing up. > > If your AI draft could have been written by anyone in your title, it is not done. Put a day in it. Put a cost in it. Then cut the throat-clearing.

That is still a LinkedIn post. It has a hook, a scene, a takeaway, a line a reader can steal. A model can hold that shape. It cannot invent the Thursday batch unless you tell it you batched on Thursday.

Why the first one happens

ChatGPT has no default access to:

how you talk on a call
who you are writing for
what you refuse to sound like
what happened this week

“Viral” is not a spec. Without a spec it fills in motivational LinkedIn: long setup, numbered nothing, emoji CTA.

The problem is not that “AI cannot write.” The problem is that empty in, empty out, and then you paste it into the composer without looking at the body.

What to put in front of the model

Keep this small. Three parts, not a prompt novel:

Who + who it is for — one line. “Founder. Writing for other founders who ship.”
The note — a scene, a quote, a belief. The Tuesday. Paste the messy version.
The shape — hook, story, takeaway, a question that is not “agree?” Short sentences. No “fast-paced world.”

Example you can steal:

> Voice: direct, short sentences, no TED-talk openers. Audience: operators who post on LinkedIn. Note: I deleted a draft that started with “in today’s fast-paced world.” I batch three posts on Thursday; I do not post daily. Write a LinkedIn post with a hook, the scene, one takeaway. No numbered tips unless they came from the note.

Then edit. If you would not say a line in a room, cut it. The model will still sneak in “journey,” “leverage,” and “game-changer.” That is your job, not a setting.

A prompt does not remember last week’s comments. For that you need a system: pillars, inbox, cadence. That argument is in ChatGPT can write posts — but it cannot build a content system (https://www.influensea.com/blog/chatgpt-can-write-posts-but-not-content-system). The checklist is in the LinkedIn content system (https://www.influensea.com/resources/linkedin-content-system).

Look at it as a post, not as a chat

Generic output survives because it looks fine in a ChatGPT panel. On LinkedIn it dies: no line breaks, a wall of tips, a CTA under “see more” that nobody opens.

Paste the draft somewhere you can see the body on the card. If the first two lines do not earn the rest, rewrite the first two lines. Do not publish from the chat window.

Influensea is built for that check: generate a LinkedIn draft and preview it as a post with a body before it goes out. It will not train a celebrity clone of your soul. It will not post 10× for you. It will stop you from shipping the before-example because you never looked at it as a post.

The free plan includes 5 LinkedIn drafts per month; no credit card.

What to do on the next draft

Do not type “write a viral LinkedIn post.”
Paste one real note from this week.
Ask for hook → scene → takeaway, in your cadence.
Preview the body. Cut anything from the before-example.
Ship. Reply in the first hour.

Generate the next draft (https://www.influensea.com/sign-up) from a note, not from a wish. If it still sounds like the before block, the note was too thin — not the model.

1a:["$","article",null,{"children":[["$","$L7",null,{"className":"mb-8 flex w-max items-center gap-1 text-muted-foreground leading-none hover:text-accent","href":"/blog","children":[["$","svg",null,{"ref":"$undefined","xmlns":"http://www.w3.org/2000/svg","width":24,"height":24,"viewBox":"0 0 24 24","fill":"none","color":"currentColor","className":"mb-0.5 h-5 w-5","children":[["$","path","0",{"d":"M15 6C15 6 9.00001 10.4189 9 12C8.99999 13.5812 15 18 15 18","stroke":"currentColor","strokeLinecap":"round","strokeLinejoin":"round","strokeWidth":"1.5","opacity":"$undefined"}]]}]," All Posts"]}],["$","h1",null,{"className":"my-2 font-bold text-5xl sm:text-6xl md:text-7xl lg:mb-6","children":"Why your ChatGPT posts suck (and how to fix them) |"}],["$","p",null,{"className":"text-lg text-secondary-foreground sm:text-xl","children":"ChatGPT LinkedIn posts sound generic because you asked for a viral post and fed it nothing. Here is the same topic as a before and after: sludge vs a scene you would actually say. Then preview the body before you ship. |"}],["$","header",null,{"className":"my-4 lg:my-6","children":["$","address",null,{"className":"mb-6 flex flex-wrap items-center justify-between gap-3 not-italic","children":[["$","p",null,{"className":"font-medium text-primary md:text-lg","children":"lekterable"}],["$","div",null,{"className":"flex items-center","children":["$undefined",["$","time",null,{"className":"mx-1 text-primary text-sm md:text-base","dateTime":"2025-11-30T07:22:53.682Z","title":"2025-11-30","children":"November 30, 2025"}],["$","div",null,{"className":"mx-1","children":" ・ "}],["$","p",null,{"className":"text-muted-foreground text-sm md:text-base","children":[5," min read"]}]]}]]}]}],["$","figure",null,{"children":["$","div",null,{"className":"relative w-full","style":{"aspectRatio":"16/9"},"children":["$","$L1d",null,{"src":"http://zip.kornel.me/u/yZ7CqS","alt":"Why Your ChatGPT Posts Suck (And How to Fix Them in 2025)","className":"h-full w-full rounded-lg object-cover","onLoad":"$undefined","onError":"$undefined","fill":true,"sizes":"(max-width: 640px) 100vw, (max-width: 1280px) 50vw, 1280px"}]}]}],["$","div",null,{"suppressHydrationWarning":true,"className":"prose prose-lg prose-li:my-0 mt-8 prose-headings:mt-10","dangerouslySetInnerHTML":{"__html":"$1e"}}],"$L1f","$L20"]}]