How to write social media captions with AI (that don't sound like AI)

AI can take a lot of the grind out of social media: turning a rough idea into a first draft, adapting one post for five networks, suggesting hashtags, or rewriting a caption that isn't landing. Used badly, it produces the same polished, empty posts everyone has learned to scroll past.

The difference isn't the model. It's what you give it and what you do with what it gives back. Here's a process that produces captions worth publishing.

What AI is good at — and what it isn't

AI is very good at structure, variation and speed. It can produce ten opening lines in seconds, reshape a 1,000-character LinkedIn post into a 250-character X post, and turn bullet points into flowing sentences.

It's bad at the things that make a post worth reading: specific facts it doesn't have, opinions, real experience and your brand's particular voice. It will fill those gaps with plausible generalities unless you supply them.

Step 1: Give it the substance, not just the topic

“Write a post about our new feature” produces a generic announcement. The more specific your input, the less generic the output. A good brief has:

• The one thing you want people to take away, in a sentence.
• A concrete detail only you know: a number, a customer quote, a before-and-after, a mistake you made.
• Who it's for — “freelance designers who invoice monthly” rather than “our audience”.
• The network it's going to, so the length and tone fit.
• What you want the reader to do next, if anything.

Compare “Announce our scheduling feature” with “Tell freelance designers that they can now schedule a month of Instagram posts in one sitting; one beta user cut her Sunday posting routine from two hours to twenty minutes.” The second brief contains a real story, so the draft will too.

Step 2: Ask for options, then choose

Don't ask for one caption. Ask for several opening lines, pick the strongest, and build from there. Choosing between options is faster and produces better results than repeatedly regenerating a single answer and hoping.

Step 3: Edit out the AI tells

Readers have become good at spotting AI-written posts. These are the giveaways worth removing before you publish:

• Openings that announce rather than say: “In today's fast-paced world…”, “Let's dive in”, “Here's the thing:”.
• Inflated vocabulary: “unlock”, “elevate”, “game-changer”, “seamless”, “revolutionise”.
• Lists of three adjectives where one precise word would do.
• Every sentence the same length and rhythm.
• Emoji at the start of every line.
• A closing question that nobody would actually answer, like “What do you think?”

Then add one thing the AI couldn't have written: a detail from your own experience, an opinion, or a line in the way you'd actually say it. That single human sentence changes how the whole post reads.

Step 4: Adapt it per network — don't copy-paste

This is where AI saves the most time. Once you have a version you're happy with, ask for a version for each network: a short, punchy one for X and Threads, a fuller one with context for LinkedIn, a hook-first caption for Instagram, a one-liner for TikTok. Then review each one — the tone shift between networks is where AI most often overshoots.

Step 5: Generate hashtags from the post, not a list

Pasting the same block of hashtags onto every post is a habit worth breaking. Hashtags should describe what this particular post is about, and the right number depends on the network — a handful on Instagram and TikTok, two or three on LinkedIn, one or two on X. AI is good at suggesting specific tags from a caption's subject; your job is to delete the ones that are too broad to matter.

Useful prompts to start from

• “Give me five opening lines for this post. Each under 80 characters. No questions, no emoji.”
• “Rewrite this LinkedIn post for X in under 240 characters. Keep the specific number.”
• “Make this caption more direct. Cut any sentence that doesn't add information.”
• “Suggest five specific hashtags for this Instagram caption. Avoid tags with tens of millions of posts.”
• “Turn these three bullet points into a 150-word LinkedIn post in the first person.”

AI captions in Postreo

Postreo puts the AI writing assistant inside the post composer instead of in a separate chat window. Describe the post in a line and it drafts the caption, suggests hashtags from what the post is about, and creates a version tuned for each network you're publishing to. You can rewrite any one version — shorter, warmer, more direct — without losing the others, and generate a matching image or short video in the same place.

Frequently asked questions

Do social networks penalise AI-written posts?

Networks rank posts on how people respond to them, not on how they were written. AI-written posts underperform when they're generic, because people don't engage with generic posts — which is why the editing step matters.

Should I disclose that a caption was written with AI?

For ordinary captions that you've edited and stand behind, most brands don't. AI-generated images and videos that look realistic are different: several networks ask you to label them, so check each network's current rules.