ChatGPT prompts for social media posts that fit each network

Short answer

A prompt that works names the goal, the audience, the facts ChatGPT may use, the network with its limit, the tone and the format of the answer. OpenAI’s own guidance says to be clear and specific, to give enough context and to refine in rounds. Below are seven templates we wrote, with the limits that Mellow Hub checks filled in: 2,200 characters on Instagram, 280 on X, 3,000 on LinkedIn. Paste one, replace the brackets, and count the result before it goes out.

Written by the Mellow team · Updated October 10, 2026, sources read October 10, 2026

What OpenAI’s guidance asks for

From OpenAI’s help article on prompt engineering and its prompt engineering guide, read on 10 October 2026:

Be clear and specific. Give the model enough context, and avoid ambiguity.
Refine in rounds. Start with a prompt, read the answer, then adjust the wording, add context or simplify.
Name the tone with adjectives. Words such as friendly, formal, humorous or serious steer it.
Structure the prompt. The guide suggests Markdown headings or XML tags to separate parts, in the order of identity, instructions, examples and context.
Show examples. It recommends input and output examples that cover a range of cases.
State the format. Say what the answer should look like.
Test it. The guide says model output is not deterministic and recommends tests that measure how a prompt behaves.

The templates below are built on those points. They are starting points to change, not a promise about results.

The six parts of a post prompt

Part	What to put in it	Why it is there
Goal	What the post should make someone understand or do	One post, one job
Audience	Who reads it and what they already know	Sets vocabulary and detail
Facts	Only true things: dates, names, numbers, prices	The model should use these, not guess
Network and limit	Which network, and its character limit	The post has to fit before it can be sent
Tone	Two or three adjectives	Keeps the voice yours
Format	How many versions, labeled how, with or without hashtags	A format you can paste from

Limits to put in the prompt

These are the numbers Mellow Hub checks, each from the network’s own documentation:

Network	Limit	The detail to include
Instagram	2,200 characters, up to 5 hashtags	Put the hashtags at the end
X	280 on X’s weighted count	A link counts as 23 whatever its length, and an emoji as 2
Threads	500 characters	An emoji counts as its UTF-8 bytes, so a pointing hand is 4
Bluesky	300 graphemes	Counted in Unicode grapheme clusters, not bytes
LinkedIn	3,000 characters, and text is required	Say that the first lines carry the point

The other networks’ rules are on the networks page.

Seven prompts

One announcement for three networks

Shorten without losing the point

A week of posts from a list of facts

Opening lines

A long text as a thread

Edit, do not rewrite

Check it with Mellow Hub before it goes out

The last one needs ChatGPT to have Hub added as an app, which OpenAI offers in developer mode on some plans: see the ChatGPT guide and can ChatGPT schedule posts.

Examples work better than adjectives

OpenAI’s guide recommends showing input and output examples. For a post, that means pasting two of your own earlier posts above the request with the line “Match the voice of these two posts”. It tells the model more than “friendly and professional” does.

Check before you publish

Do not rely on ChatGPT to count. Paste the result into the free character counter, which counts X and Threads the way those networks do, or into the post checker, which applies Hub’s input rules for each destination. The hashtag counter is there for Instagram’s limit of 5.

Hub is paid: plans start at $19 a month (Solo), after a 7-day trial with a card on file that costs nothing if you cancel before it ends, and a monthly plan's first month is 20% off. Publishing is unlimited on every plan, under a fair-use ceiling that stops automated floods (900 publications a month on Solo). One publication is one post on one account, so a post sent to five networks counts as five. There are no automatic overage charges.

Questions people also ask

How do I make ChatGPT write in my voice?

Describe the tone in two or three adjectives, as OpenAI’s help article suggests, and paste two of your own earlier posts as examples, which its prompt engineering guide recommends. Ask it to match the voice of those posts and to use only the facts you give it.

How do I know a post fits a network’s limit?

Count it with a tool that counts the way the network does. X weights characters and counts a link as 23, and Threads counts an emoji as its UTF-8 bytes, so a plain character count can be wrong. Hub’s free character counter shows the numbers Hub’s own check uses.

Do I have to label a post ChatGPT wrote as AI?

The labeling rules we read are TikTok’s, which defines content made with AI as images, video and audio, and YouTube’s, which covers realistic altered or synthetic content in video. Neither describes a written caption as something to label, but check the rules of each network you post on, and of any picture or video the post carries.

Can ChatGPT post for me once the prompt works?

Not by itself. It can write the post, and with a custom MCP app added in developer mode, on the plans OpenAI allows, it can call a service such as Mellow Hub that checks and publishes. Otherwise you paste the text into the network or a scheduler.

Why ask it to use only the facts I give it?

Because the post goes out under your name and the network will not check its claims. OpenAI’s guide says a model’s output is not deterministic and should be tested, so a prompt that fixes the facts, and tells the model to ask when one is missing, gives you something to check the answer against.

Sources

Other companies’ features change. Every statement about them rests on these pages, as they read on the date above.