Writing one good long-form piece is hard. Writing ten distinct posts to promote it is harder, and that is where most creators quit. This workflow takes a single source (a blog post, podcast episode, or YouTube video) and turns it into ten or more platform-tailored posts in one 60-90 minute session, each in the native voice of its platform, all linking back to the original.
TL;DR
- Run the workflow within 48 hours of publishing, while the source is still vivid. You write sharper hooks that way.
- Four steps: extract durable assets (insights, quotes, frameworks, contrarian takes, stats), pair each asset with a platform, generate native-voice posts, schedule over two to three weeks.
- A general LLM (Claude Opus 4.7 or GPT-5.5) does the text extraction and drafting. Both ship a 1M-token context as of June 2026, so a full transcript fits in one paste.
- For video-first sources, pair the LLM with a clipper such as Opus Clip (free tier: 60 credits/month, paid from $15/month) to cut the actual clips.
- Edit every draft. AI scaffolds; the lines that convert are the ones only you can write.
What this covers
A repeatable four-step process: extract durable assets from your long-form, pair each asset with a platform, generate native-voice posts, and schedule them over two to three weeks. The output is a content calendar fully populated from one creative effort, plus a link-back funnel pointing every post at the original.
Who this is for
- Solo creators publishing a weekly podcast or YouTube who struggle to fill social calendars between drops.
- Brand teams with one main channel (a blog or newsletter) and four secondary channels (X, LinkedIn, Instagram, YouTube Shorts).
- Consultants whose long-form work is research-heavy and rarely surfaces beyond LinkedIn.
Less useful for hobbyist posters who only have time for one channel anyway.
When to reach for it
After publishing any long-form piece over 1,500 words or 30 minutes of video. Run the workflow within 48 hours while the source is fresh; you write better hooks when the content is still vivid. Also use it before a launch week to bank a backlog of related posts.
Platform format limits (verified June 2026)
Get these right before you generate, so the AI writes to the real constraint instead of overshooting. As of June 2026:
| Platform | Hard limit | Practical sweet spot |
|---|---|---|
| X (free account) | 280 chars per post | 200-250 chars; thread for more |
| X (Premium, $8/mo web) | 25,000 chars long-form | First 280 show before “Show more” |
| LinkedIn post | 3,000 chars | 1,300-1,900 chars (~300-400 words) |
| Instagram carousel | 20 slides in-app (10 via most schedulers) | 5-7 slides; completion drops after 7 |
| Reels / Shorts script | n/a (length is video) | 30-60 seconds spoken |
| Newsletter intro | n/a | 2-3 short paragraphs plus the link |
Two counting gotchas on X: every URL costs exactly 23 characters (t.co wrapping), and most emoji cost 2. Tell the AI to budget for the link.
Pick your tools
You only need one general LLM for the text work. Add a clipper only if your source is video.
| Job | Tool | June 2026 pricing | Notes |
|---|---|---|---|
| Extract + draft text | Claude Opus 4.7 (Pro $20/mo) | $0 limited / $20 Pro | 1M-token context; strong at voice-matching |
| Extract + draft text | ChatGPT GPT-5.5 (Plus $20/mo) | $0 with ads / $20 Plus | Plus in-app context ~320 pages; good thread structure |
| Cut video clips | Opus Clip | Free 60 credits/mo, Starter $15/mo, Pro $29/mo | 1 credit = 1 min of source video; paid removes watermark |
| Multi-platform scheduling | Buffer / native schedulers | Free tiers exist | Hold final drafts here, not the AI output |
Either LLM works. Claude tends to hold a voice profile more consistently across many posts; GPT-5.5 tends to structure threads cleanly. Use whichever you already pay for. See Claude vs ChatGPT for long documents if you are deciding.
Step by step
- Anchor the source. If the long-form is a podcast you have not recorded yet, anchor the recording first with an AI podcast episode outline (hook, five segments, transitions) so the repurposed clips already align to a structured arc. For existing pieces, paste the full text or transcript into the AI. A full hour-long transcript fits inside the 1M-token context of Opus 4.7 or Gemini 3.1 Pro in a single paste.
- Extract durable assets. Use the extraction prompt below to pull insights, verbatim quotes, frameworks, a contrarian take, and supporting stats in one pass.
- Pair extracts with platforms. Insights to LinkedIn posts; quotes to X singles or thread tweets; frameworks to Instagram carousels or LinkedIn documents; contrarian take to an X thread or LinkedIn debate post; supporting stats to the newsletter intro.
- Generate native-voice posts. Prompt for each: “Write a [platform] post using [extract]. Voice: [your voice notes]. Constraints: [platform limit from the table]. Link back to [URL]. Give me three variants.”
- Edit, do not approve raw. AI gives a first draft per platform; cut anything that sounds like AI, add one sentence only you could write, and shorten the hook. Five minutes per post is the right pace.
- Schedule over 2-3 weeks. Spread posts to avoid cannibalizing your own attention. Different platforms can run simultaneously; same-platform posts need 48-72 hours apart.
- Link back consistently. Every post names the source: “Full breakdown on the blog,” “Episode 47 of the podcast,” “Wrote this up in detail, link in comments.” This is the SEO and funnel value.
Asset extraction prompt template
Source: [paste long-form text or transcript]
Extract:
- 3 STANDALONE INSIGHTS (each a complete idea, no context needed)
- 5 QUOTE CANDIDATES (verbatim sentences under 30 words)
- 2 FRAMEWORKS (numbered steps or named model)
- 1 CONTRARIAN TAKE (opinion that pushes against consensus)
- 3 SUPPORTING STATS (specific numbers, with the source line they came from)
For each item, name the platform it fits best and why.
Quote candidates must be word-for-word from the source. Do not paraphrase.First-run exercise
- Pick your last long-form piece. Run the extraction prompt.
- Generate three posts targeting your most active platform only.
- Post the strongest of the three. Compare engagement to your usual post on that platform.
- Adjust your voice notes for next time based on what landed.
Quality check
- Do the posts sound like you, or like ChatGPT? AI defaults to em-dashes, “imagine if,” and three-clause sentences. Cut them.
- Does each post stand on its own without the source? A reader who sees only the post should still get value.
- Are you actually linking back, or just hoping? Every post should name the source and offer a click path.
- Did you over-pack a single platform? Two LinkedIn posts in one week is right; six is spammy.
How to reuse this workflow
- Save the extraction prompt and the per-platform generation prompts as separate templates. Mix and match per piece.
- After four to six repurposes you will see which extract type drives the most engagement on each platform; weight future runs toward those.
- Build a “best posts” library. The strongest posts are also your strongest voice samples to feed the AI next time.
Recommended workflow at a glance
Long-form source, AI extracts (insights, quotes, frameworks, contrarian, stats), platform pairing, generate posts with voice notes, human edit pass, schedule across two to three weeks, link back consistently. Time investment: 60-90 minutes once per long-form, replacing five to eight hours of ad-hoc social posting.
Common mistakes
- Cross-posting identical text across platforms. Audiences notice and you waste the multi-platform reach.
- Not adapting voice per platform. LinkedIn formality on X reads as a bot; X casualness on LinkedIn reads as low-effort.
- Skipping the link back to long-form. The whole point is the funnel; without the link the work does not compound.
- Trying to repurpose every long-form. Some pieces do not hold ten distinct insights; forcing extracts produces filler.
- Posting all of it in 48 hours. You exhaust your own audience, and the long-form gets one weekend of attention instead of two weeks.
- Letting AI write the final voice. AI scaffolds; you add the lines a model cannot generate because they require lived experience.
FAQ
- Can I repurpose someone else’s long-form?: Only with permission and clear attribution. Otherwise you are republishing, not repurposing.
- How many posts is too many?: Above 12 per long-form, quality drops because you are stretching insights too thin. Eight to twelve is the sweet spot.
- Does AI extract quotes accurately?: Mostly, but verify every quote against the source word-for-word. AI occasionally paraphrases when asked for “quotes.” Paste it the literal text and re-prompt with “verbatim only.”
- Which model should I use?: Either Claude Opus 4.7 or GPT-5.5 works for the text. As of June 2026 both ship a 1M-token context, so a full transcript fits in one paste. Claude tends to hold a voice profile more consistently; GPT-5.5 tends to structure threads more cleanly.
- Should I repurpose backwards too?: Yes. A strong tweet thread can become a blog post. The workflow is platform-agnostic; long to short is just more common.
- What about video?: Same logic plus an editing step. Extract 5-7 key moments by timestamp from the transcript, then cut those with a clipper like Opus Clip (free for 60 minutes of source video per month). Pair each clip with a platform-native vertical or horizontal format.
- How long should this take per piece?: 60-90 minutes the first time; 30-45 minutes once you have templates. Much faster and quality drops; much slower means you are over-editing.
Related
- Content calendar prompts
- Cross-platform repurpose
- AI writes short-form video hooks
- AI writes an X (Twitter) thread
- AI writes an Instagram Carousel script
- AI creator brand tutorial
Tags: #Tutorial #Content creation #Repurpose #Multi-platform