TL;DR
One blog post can feed five surfaces, but copying the same text into all of them fails everywhere. The job is to find the one idea per platform the post supports, then write a native version that fits each surface’s length window, hook style, and CTA norm. Feed the AI a reference example in the target voice for every platform, run two passes (facts first, polish second), and stagger publishing across 2-3 days so the versions don’t cannibalise each other’s reach.
The task
You published a blog post and you want maximum reach without writing four more pieces from scratch. The wrong output is “the same content in four formats.” The right output is the one idea per platform the blog post supports, rewritten with the vocabulary, rhythm, and hook that fits that surface — and sized to where each platform truncates.
When AI helps and when it does not
AI is excellent at format translation, hook variation, and length compression. It is poor at native voice. What reads as native on X reads as alien on LinkedIn, and vice versa. Give the AI a reference post in the target voice for each platform; otherwise you ship a flat, translated version that performs worse than the blog post on every surface.
As of June 2026, the models worth using for this are Claude Sonnet 4.6 (strong at matching a pasted reference voice and following multi-platform instructions in one shot), GPT-5.5 (good hook variation, slightly more eager to add exclamation marks), and Gemini 3.1 Pro (all three carry a 1M-token context, so a long-form post plus four reference examples fits with room to spare). Voice-matching beats raw reasoning here, so the reference example matters more than the model choice.
Platform specs that decide the format (June 2026)
Repurposing fails when the draft ignores where each surface truncates or caps. These are the numbers to write to:
| Platform | Hard cap | Where it truncates / sweet spot | Format that wins |
|---|---|---|---|
| X (free) | 280 chars per post | Threads have no length cap; each post stays 280 | 5-8 post thread; first post is the hook |
| X (Premium long post) | 25,000 chars (~4,500 words) | Only first ~280 show before “Show more” | Use only if the idea needs one unbroken argument |
| 3,000 chars | ”…see more” cuts around 200-210 chars on desktop, ~140 on mobile | 150-250 word post; hook lands in the first 2 lines | |
| Short video (Reels) | 90s standard (longer clips suppressed for new viewers) | 21-34s outperforms on TikTok; ~55s on Shorts | Hook in first 3 seconds + one payoff |
| YouTube Shorts | 60s hard | Aim ~55s to maximise watch time | Same as above |
| Instagram carousel | 20 slides | Completion stays high at 5-8 slides | Slide titles only, one idea per slide |
| LinkedIn carousel (document post) | 300 pages / 100 MB PDF | 3-10 slides outperform | Export to PDF; one point per page |
Sources: X character limits 2026, LinkedIn limits 2026.
What to feed the AI
- The full blog post
- The one-sentence thesis (your own, not the AI’s distillation)
- Target platforms with their typical reader
- A reference example you like for each platform (your own past hit, a competitor’s, an author’s voice you admire)
- Brand voice constraints (no exclamation marks, no ”🧵”, no engagement bait)
- The CTA per platform (drive to the blog, grow the list, or just impressions)
Copy-ready prompt
Repurpose the following blog post into platform-native assets.
Thesis (one sentence): [line]
Brand voice constraints: [list]
For each target platform, here is the format, the reader, and a
reference example to match the voice of:
Platform: X (Twitter)
Reader: [who]
Format: 6-post thread; each post under 280 characters; post 1 is the hook
Reference (paste): "[200-300 words from a post that performed well]"
CTA: [drive to blog / no CTA]
Platform: LinkedIn
Reader: [who]
Format: 180-word post; the hook must land in the first 210 characters
(that is all that shows before "see more")
Reference: "[paste]"
CTA: [comment for link / DM]
Platform: short video (TikTok / Shorts / Reels)
Reader: [who]
Format: ~45-second script; hook in the first 3 seconds, one clear payoff
Reference: "[paste]"
Platform: newsletter section
Reader: [who, our subscriber profile]
Format: 150-word intro with a click-out to the full post
Reference: "[paste]"
Blog post:
"""
[paste]
"""
Return one platform-native draft per platform. Do not just shorten
the blog — re-hook for each surface. Mark anywhere the original
argument was simplified, so I can verify it stayed honest.For visual-first platforms, add: “Also write a 6-slide Instagram carousel script with slide titles only, no walls of text, one idea per slide.”
Two-pass workflow
- Pass one — facts. Ask the model to repurpose without polishing tone. Check that the thesis survived compression and no simplification changed the argument.
- Pass two — voice. Now ask it to match the pasted reference for each platform. Keeping these passes separate stops the model from inventing a detail while it is busy polishing a sentence.
How to check the output is usable
- Each version reads as native (read it aloud in the target voice)
- The hook is platform-appropriate: curiosity for X, a story or stance for LinkedIn, a visual or claim in the first 3 seconds for video
- The thesis survives compression and none of the simplifications are wrong
- The LinkedIn hook fits the first ~210 characters; the X opener works as a standalone post
- CTAs match the platform norm (no “link in bio” on LinkedIn, no ”🧵” on Threads)
- A reader who sees only one version still gets the idea
Common mistakes
- Identical text across platforms. Performs poorly everywhere
- LinkedIn cadence on X and vice versa. The surfaces reward different rhythms
- Burying the LinkedIn hook past character 210, where “see more” hides it
- Lossy simplifications that change the argument. Flag and verify
- Engagement-bait CTAs (“agree? RT if yes!”). Short-term lift, long-term reach penalty
- Skipping the reference example, which leaves the AI defaulting to generic AI voice
- Publishing all five the same day, so the versions split one audience instead of compounding
FAQ
- Evergreen vs newsy posts? Evergreen posts repurpose 4-5 times across a year; newsy posts get one wave, so front-load them.
- Should I publish all at once? No. Stagger by 2-3 days. Same-day blasts cannibalise each other’s reach and split the same audience.
- Which model is best? Voice-matching matters more than raw reasoning, so any of Claude Sonnet 4.6, GPT-5.5, or Gemini 3.1 Pro works as of June 2026. The reference example you paste moves the result more than the model does.
- How do I handle accidental contradictions? Use the “simplification notes” the prompt asks for. If a claim was softened on one platform but kept strong on another, fix it before publishing.
- Should I use X long-form posts instead of a thread? Usually no. Most readers stop at the 280-character truncation, so a well-structured thread of native posts often out-reaches one long-form post with the same content.
Related
- AI content repurpose tutorial — end-to-end workflow
- X thread prompts — focused thread variants
- X thread AI — workflow for threads from scratch
- Article rewrite — tone shift on the same content
- Newsletter AI — newsletter-specific repurposing
- Viral shorts story arc prompts — short-form video arc
Tags: #AI writing #Content creation