AI Prompts to Repurpose Blog Posts for Social Media

You spent four hours on a blog post. It lives on exactly one URL while six other platforms — LinkedIn, X, Instagram, email, Telegram, Reddit — sit empty. The standard fix everyone repeats is "just use ChatGPT prompts." So you do, and two things go wrong: you burn 45 to 90 minutes hopping between a dozen prompts per post, and the output sounds like every other AI-written post on the timeline. Here are AI prompts to repurpose blog posts for social media that work as a system, not a list — seven prompts that run in a deliberate order, each one feeding the next, so the output stays coherent and sounds like you. The short version: run one extraction prompt first to pull the spine of the post, then feed that spine into six platform-specific prompts. Skip the extraction step and every downstream output drifts generic. Most of the guides ranking for this term are listicles of 30 to 50 prompts with no sequencing and no explanation of why each prompt is phrased the way it is. This is the opposite. Seven prompts, in order, each one explained — and an honest note at the end about when running them by hand stops being worth it. If you want the conceptual frame first, the deeper read on what content repurposing actually means sits at voxplit.com/blog/what-is-content-repurposing.

Almost everyone skips this step, and it is the reason most AI-repurposed content reads as generic. If you paste a full 1,400-word article into a "rewrite this as a LinkedIn post" prompt, the model averages the whole thing into mush — it has no idea which claim matters most, so it hedges everything equally. Extraction fixes that by forcing the model to pick the spine before it writes a single platform post. Run this one prompt first. Its only job is to surface three lines — the sharpest claim, the best concrete proof, the one action a reader can take this week. Every platform prompt below uses those three lines as its context block instead of the full article. That single decision is what keeps a thread, a carousel, and an email all pointing at the same argument instead of wandering off in three directions. Prompt 1 — Extraction: You are a content strategist. I'm going to paste a blog post below. Before we repurpose it, extract: (1) the single sharpest claim or insight, (2) the best concrete example or statistic, (3) the one actionable takeaway a reader could apply this week. Do not rewrite yet. Just extract those three lines. [PASTE BLOG POST] Why "do not rewrite yet" is in there: without it, the model jumps straight to drafting and you lose the clean spine. The instruction forces a pause. Save the three lines it returns — that block is the raw material for everything that follows.

LinkedIn rewards a specific opening line and punishes the obvious AI tells. The feed dwell-time algorithm reads short paragraphs and a genuine question at the end as signals of quality; it reads "I'm excited to share" and a wall of hashtags as spam. So the prompt below bans the clichés explicitly and caps paragraph length, because those are the exact places generic output gives itself away. Notice the prompt asks for your job title, niche, and audience in the first line. That is not optional decoration — it is the difference between a post that sounds like a real practitioner and one that sounds like a template. Fill the brackets in every time. Prompt 2 — LinkedIn: You are a LinkedIn ghostwriter for [describe yourself: job title, niche, audience]. Using the three extracted lines above, write one LinkedIn post. Rules: open with a specific, non-clickbait claim in the first line (no "I'm excited to share"). Use short paragraphs, max 3 lines each. End with one open question that invites replies. Tone: direct, professional, first-person. Length: 120-180 words. Do not add hashtags. This prompt produces a text post. If your blog post is a how-to or a framework, it will often convert better as a swipeable deck instead — the step-by-step LinkedIn carousel method at voxplit.com/blog/blog-post-to-linkedin-carousel covers when to choose the carousel and how to map each idea to a slide.

The thread format works on X for one structural reason: each tweet has to survive on its own. Someone sees tweet 4 of 7 because a friend quoted it, with zero context — if that tweet still makes a point, the thread gets reshared; if it only works as a continuation of the one above, it dies. So the prompt forces standalone readability per tweet and a hard character ceiling, which doubles as a compression test. If an idea won't fit in 280 characters, you haven't sharpened it enough yet. Prompt 3 — Twitter/X thread: Using the three extracted lines, write a Twitter/X thread. Format: numbered (1/7, 2/7, etc.). Tweet 1: hook — a single surprising or contrarian claim under 220 characters, no em-dashes. Tweets 2-6: one concrete idea per tweet, each standalone readable. Tweet 7: one CTA — either a question or "full post at [link]". Hard limit: 280 characters per tweet. No emojis. Voice: [your tone — e.g. "direct, slightly skeptical, no hype"]. The "no em-dashes" rule in the hook is deliberate: the em-dash is one of the loudest AI tells on X right now, and it shows up most in opening lines. Banning it from the hook alone cleans up the part of the thread that gets seen most. For sequencing, publishing relative to the original post, and which blog posts convert to threads at all, the full Twitter/X thread workflow lives at voxplit.com/blog/blog-post-to-twitter-x-thread.

These four group together because the pacing logic is the same: each prompt names the platform's native shape, then constrains length and tone so the model can't default to generic. Run them in sequence after the extraction prompt, the same as the first two. Email. The newsletter blurb leads with one personal line and ends with a clear link CTA, because subscribers opened the email expecting your voice, not a press release. The "letter to a smart friend" framing in the prompt is what keeps it from sounding like a content marketing blast. Prompt 4 — Email newsletter blurb: Write an email newsletter section for my subscribers, based on the blog post below. The section should: start with one personal sentence about why I wrote this (1-2 lines), give readers 2-3 key takeaways in plain language (not bullet points — short paragraphs), and end with a link CTA: "Read the full post here: [URL]". Tone: like a letter to a smart friend who already knows my work. Max 180 words. No subject line needed — I'll write that separately. [PASTE BLOG POST OR EXTRACTED LINES] Instagram. Carousels are read as large text on a phone, so the prompt caps headlines at 7 words and asks for mobile-friendly body copy. The hook slide does the same job the X hook tweet does — stop the thumb. Prompt 5 — Instagram carousel outline: You are an Instagram content strategist. Using the blog post below, write a 7-slide carousel outline for Instagram. For each slide provide: a bold slide headline (max 7 words, readable as large text) and 1-2 supporting sentences (body copy, mobile-friendly length). Slide 1: hook — a bold claim or provocative question. Slides 2-6: one idea each. Slide 7: CTA — "Save this" or "Send this to [specific person]." Do not write "Slide X:" in the output — just number each slide. Tone: [your brand tone]. [PASTE BLOG POST] Telegram. Short-form channels want standalone posts with different angles, not one chopped-up article. The prompt asks for three distinct takes so you get a week of posts from one source. Prompt 6 — Telegram / short-form: Take the blog post below and write 3 short posts for a Telegram channel. Each should be 100-180 words, standalone, and use a different angle: (1) a hook/provocation that surfaces the main tension, (2) a practical tip or mini-checklist a reader can use today, (3) a personal observation or opinion tied to the topic. No hashtags. No "in today's post." Each post gets a blank line between paragraphs. Conversational but expert tone. [PASTE BLOG POST] Reddit. The one platform where marketing language gets you downvoted and removed. The prompt strips markdown headers, bans the brand mention, and asks for a real question — because Reddit rewards peer-to-peer posts and punishes anything that smells like promotion. Prompt 7 — Reddit-style community post: Write a Reddit post based on the blog post below. Format: a plain-text post with no markdown headers. Start with a direct question or observation that the community would actually engage with — not a title that sounds like marketing. Write 150-200 words of body copy that shares the insight without selling anything. End with an open question to the community. Tone: peer-to-peer, self-aware, no hype. Do not mention [brand name] in the post itself. [PASTE BLOG POST]

Once you have run the seven prompts a few times, you will notice the same three rules carry every platform. Get these right and even a mediocre platform prompt produces usable output. Get them wrong and the best prompt in the world returns filler. Rule one: always paste the source text, not just the topic. "Write a LinkedIn post about content repurposing" gives the model nothing specific to work with, so it returns the average of everything it has read on the topic — which is exactly the generic output you are trying to avoid. Pasting the actual post grounds every line in your real argument and your real examples. Rule two: always specify your audience and tone. The bracketed [describe yourself] and [your tone] fields are not filler. A post written for "early-stage SaaS founders" reads nothing like one written for "yoga studio owners," and the model can only hit the right register if you name it. Skip this and everything defaults to the flat, audience-less voice that screams AI. Rule three: use the extraction prompt output as the context block, not the full article. After Prompt 1, feed the three-line spine into the platform prompts wherever you can, instead of re-pasting 1,400 words each time. The spine keeps every output anchored to the same core argument, which is how a thread, an email, and a carousel end up feeling like one coherent campaign instead of three unrelated drafts. The deeper mechanics of avoiding generic AI output are covered at voxplit.com/blog/ai-for-social-media-posts.

Here is the friction nobody in the listicles names. Running all seven prompts by hand — extraction, then six platform adaptations, each with bracket fields filled in and an editing pass on top — takes 45 to 90 minutes per blog post. That is fine for one post. It is not fine four times a month, because the cost compounds: every published piece adds another hour-plus of prompt-hopping, and the week you skip it is the week your six other platforms go quiet. The prompts above are the right mental model. Extract the spine, adapt per platform, preserve your voice, never paste just the topic. That logic is sound. The problem is purely the implementation — you are the one copy-pasting the same source into a dozen windows, re-filling the same brackets, and stitching the outputs back together by hand. That is the exact inefficiency Voxplit removes. It runs the same logic these seven prompts describe — extract, adapt per platform, hold the voice — except you paste the source once and all seven platforms adapt simultaneously, in your existing writing voice rather than a generic template. No prompt-hopping, no re-filling brackets, no stitching. The honest comparison of where the manual-prompt approach still makes sense and where a dedicated tool wins is laid out in Voxplit vs. ChatGPT for content work at voxplit.com/blog/voxplit-vs-chatgpt. Start with one blog post this week. Run Prompt 1 to pull the spine, then run two or three platform prompts by hand so you feel the system working — and feel the time it costs. Once you know the shape of the workflow, the walkthrough of adapting one blog post for four platforms at voxplit.com/blog/from-blog-post-to-4-platforms shows what it looks like when the chaining is handled for you. The prompts teach you the logic. The decision about whether to keep running them manually is yours.