How to Repurpose One Blog Post Into a Month of Social Posts

You have the article. Two thousand words, a week of work, one tweet announcing it. There is a month of social content inside that file, and the advice you have read stops at “break it into pieces”.

This is the mechanic: one reading pass that marks five kinds of thing, a table saying which kind becomes which format, and the whole thing run end to end on a real post of ours so you can count the output rather than take my word for it.

Why ‘turn one post into twenty’ usually fails

Because people slice by paragraph. An article has six sections, so you write six posts, then split the long ones until you have twenty. Every post is a shrunken section, and a shrunken section is a summary of something nobody has read.

Two things go wrong at once. A section is a container, not an idea: it holds a claim, two figures, a procedure and an aside, and the post you cut from it inherits all four and lands on none. And slicing by section produces posts in the article’s order, which is the order of an argument. A feed is not reading an argument. It sees one post, in isolation, next to a photo of somebody’s lunch.

The unit is the idea, not the section. One idea that somebody could disagree with, or one number they could write down, or one procedure they could follow. Get the unit right and the format question answers itself, which is why repurposing is the second of the automation workflows worth building, after the publishing queue.

How to repurpose a blog post into social posts: the extraction pass

To repurpose a blog post into social posts, read the article once, marking as you go. You are not writing anything in this pass. You are marking five kinds of extractable unit, and the only decision at each mark is which of the five it is.

• Claims. A sentence somebody could argue with. Not “reporting matters” but “a report that arrives on its own gets read, and one somebody has to assemble gets skipped”. If nobody could disagree, it is not a claim, it is a transition.
• Numbers. Any data point: a figure, limit, price, interval or threshold. Mark the qualifier with it: a floor and a ceiling are different posts, and a monthly allowance and a per-upload maximum are different facts that look identical once the unit is out of its paragraph.
• Steps. An ordered procedure, two or more stages, where the order is the content. A list of unordered tips is not a step sequence. It is several claims in a row.
• Objections. Anywhere the article says the common view is wrong, or draws a line the consensus misses. These are the most valuable units in the file and the ones writers forget they wrote.
• Quotable lines. A pull quote, meaning a sentence that works with nothing around it. Test it by reading it aloud with no setup. If it needs the sentence before it, it is not quotable, it is a conclusion.

Three rules make the pass produce something usable. One pass only. Going back to re-read is how you end up with notes about the article instead of units out of it. Mark, do not draft. The moment you start writing a caption you stop reading, and the back half of the article yields nothing. A unit must stand without its neighbours. If you cannot read the marked sentence on its own and get something, leave it unmarked, however good the paragraph is.

Expect the yield to be lopsided. Claims and quotable lines cluster in the opening and closing sections, numbers and steps sit in the middle, and objections turn up wherever the writer got annoyed. Marking this one took twenty minutes, and it is the only part of this that cannot be handed to anything else.

Keep the marked units in one list with their type, nothing else. Captions come later, from your own social media prompt templates, one template per unit type.

Mapping units to formats

Each unit type has a default format and a default network. Defaults, not rules, and the point of having them is that you stop deciding.

Unit	Default format	Where it goes first	The constraint that shapes it
Claim	Opinion post, text only	LinkedIn, then Threads	LinkedIn publishes text organically. Threads caps a post at 500 characters, so the claim has to be the whole post
Number	Quote card, one figure per card	Instagram, Pinterest, X	An X post takes up to 4 photos, or 1 GIF, or 1 video. Pinterest publishes image or video pins only, with no multi-image pin, so one card is one pin
Steps	Carousel, or a thread	LinkedIn or Instagram for the carousel, X for the thread	An Instagram carousel runs up to 20 items and crops every item to the first image’s aspect ratio, default 1:1. A swipeable LinkedIn card stack publishes organically as a document post carrying a PDF
Objection	Myth post: the wrong view, then the line	LinkedIn, then X	Needs two beats in one post, so it wants the longer limit. On Threads it has to lose the setup
Quotable line	Graphic, the line and nothing else	Instagram, Pinterest	A single image, so one line per card. An X post can carry up to 4 of them at once, and Pinterest takes one per pin

Two of those constraints decide more than they look like they should. The first is the Instagram crop: a deck drawn at 4:5 behind a square cover gets trimmed on every slide except the cover. Pick the shape, then draw the cover in it.

The second is LinkedIn. Its Posts API reference states “Organic carousel is currently not supported” beside its own carousel sample request, and keeps the Carousel format for sponsored content. The organic route for a card stack is a document post carrying a PDF, which the feed renders as something you swipe. That is what every tool selling LinkedIn carousel scheduling is doing, and it changes the artefact you have to produce: a PDF of slides, not a folder of images.

One mapping rule that is not about any platform. A unit gets one post. If you find yourself writing a second post off the same marked sentence, you have run out of units and started padding, and padding is what the reader recognises as repurposing.

A worked example

Here is the post I used as the worked example, one of ours: “What to Automate in Social Media (and What Never to)”, just under 2,500 words by the gate’s own word count, six sections, a decision rule and a three-tier table. I ran the pass on 23 October 2026 and marked 30 units. The breakdown, and what each one became:

Claims (8)

1. The two-list format breaks in week two, when a task nobody listed arrives. LinkedIn opinion post
2. Publishing through an approved API is ordinary. Manufacturing engagement is not. LinkedIn opinion post
3. A review step that publishes by default after a delay is not a review step. LinkedIn opinion post
4. Routing an inbound message is safe. Replying to it is not, and the line does not move. LinkedIn opinion post
5. Automating the drafting automates the blank page, not the publishing. Threads post, 500 characters
6. Per-platform formatting looks like craft and is a lookup table. Discarded: the three format numbers below say it better
7. Nothing graduates off the never-automate list. Threads post
8. A task that produced a visible mistake goes back to review for a quarter. Merged with the quarterly-review steps below

Numbers (8)

1. Threads caps a post at 500 characters. Quote card
2. One X post takes up to 4 photos, or 1 GIF, or 1 video. Quote card
3. An Instagram carousel runs up to 20 items, all cropped to the first image’s ratio. Quote card
4. Facebook schedules a Page post between 10 minutes and 30 days out. Discarded: too dry to carry a post alone
5. An Instagram media container expires if it is not published inside 24 hours. Quote card, and the best of the five
6. Review all of the first 50 outputs from a new generator. Merged with 7
7. Drop to one in five once the error rate holds under roughly one in twenty. Merged with 6: one card, both thresholds
8. Two quarters under the threshold before a task graduates. Quote card

Step sequences (3)

1. The gate, then three questions in order, then three tiers. LinkedIn carousel, 6 slides, and an X thread
2. Sampling: all of the first 50, record the rate, one in five at random, revert on an error until 50 come back clean. Instagram carousel, 5 slides
3. The quarterly review: rate and cost per item, graduate after two clean quarters, demote on any visible mistake. LinkedIn carousel, 4 slides, with claim 8 as the closing slide

Objections (6)

1. “Never automate replies” is rounded wrong: X allows a narrow, opted-in case. Myth post, LinkedIn
2. Meta’s spam policy is frequency-based rather than a ban on automation. Merged with 3
3. Staying under a published frequency is no defence. Merged with 2: one myth post, both halves
4. Cross-posting identical copy is ordinary, except across X accounts you operate. Myth post, X
5. A reach penalty is the wrong explanation for why template replies fail. Discarded: needs an argument that lives in a different article
6. A queue can run perfectly and publish nothing anyone can see. Myth post, LinkedIn, the strongest unit in the file

Quotable lines (5)

1. “A complaint answered by a machine is not a 2% error, it is a screenshot.” Graphic
2. “You have automated the blank page, not the publishing.” Graphic
3. “Your queue has no idea what happened this morning.” Graphic
4. “An alert channel nobody owns is a queue nobody is watching with extra steps.” Graphic
5. “Most bad automation decisions are not made by people who got the answer wrong.” Discarded: the half that lands is the half too long for a card

The arithmetic: 30 units marked, 4 discarded, 3 folded into the unit beside them, and the first step sequence doubling as both a carousel and a thread. That is 24 posts out of one article: 10 text, 10 single images, 3 carousels, 1 thread. It is 24 rather than 30 because the discipline is in the throwing away.

Two observations worth more than the total. The objections produced the best posts and there were only six of them in this one, which is the number I would look at first on the next article rather than its length. All four discards read well inside their paragraph and said nothing outside it.

Scheduling the set without repeating yourself

Twenty-four posts at five a week runs just under five weeks. The order is where this goes wrong, because the tempting order is the one you extracted them in, and that publishes every claim in week one.

Three spacing rules, in order of how much they matter.

• Never two of the same unit type in a row on one network. A claim, then a number, then an objection. Readers register the pattern long before they register the subject, and three claims in three days reads as one person with one idea.
• Keep merged and sibling units a fortnight apart. The two posts drawn from the same section are the ones most likely to read as the same post. Sort the set by source section to find the collisions.
• One carousel a week, maximum. Three carousels is three weeks. They are the heaviest things to make and the most obviously templated when they arrive together.

Vary at the unit level, not by rewriting captions. A claim as an opinion post and the same claim as a question are the same post twice. Give about a third of the set a CTA back to the article and let the rest stand alone, because a post that always ends in a link reads as an advert for a post. On Threads that habit has a hard edge: a text-only post carrying more than five links fails at media creation, which an image or carousel post does not.

Then the last step: after the month, promote the winners rather than reposting the set. Pull the numbers once, which is what the weekly report template is for, and move the three or four posts that actually worked into a recycle queue with a minimum interval and a per-item kill switch. That is a different mechanism from reposting, and the evergreen queue covers how the interval and the switch are set. Reposting a whole set immediately is the one move that teaches an audience to ignore you, and it makes the extraction look like the problem when the scheduling was.

Where to automate it

Split the work by whether the output depends on who did it.

Machine work. Transcription, when the source is a recording rather than a post. Variant generation, once a human has chosen the unit and written the first version. Resizing and reshaping to each network’s canvas. Building the PDF a LinkedIn card stack needs. Scheduling and publishing the set. Pulling the numbers afterwards. All of it is rule-following, and content repurposing software is the shortlist for the transformation half.

Human work. Choosing which sentences are units, which is the whole pass. Writing the opinion, because a model will write a confident claim nobody at your company holds. Deciding the order. Deciding what to throw away, which is the only reason this produces 24 posts rather than 30 mediocre ones.

On the publishing end a scheduler is enough. SchedPilot’s live MCP server exposes a dedicated LinkedIn carousel call, which we confirmed by reading the tool’s own description: 2 to 20 image slides in order, a document title of 100 characters or fewer, any caption burned onto a slide capped at 200, no video, and it states that it turns the slides into the PDF document LinkedIn needs when it publishes. One shape rule comes with that: it crops any slide you put text on to 4:5 portrait, so draw the whole deck at 4:5 if a single slide carries a caption. It publishes slides you supply and does not design them: each slide is an image you upload first, so the deck is still yours. Its site advertises a built-in editor for short clips with captions and text overlays, and documents no clip detection and no transcription, so the transformation happens somewhere else first. Its post call takes an X thread as a main post plus ordered replies, the shape a step sequence wants anyway. If the trigger has to come from your CMS, that is where Zapier, Make or n8n earns its place, and the wider picture of a social media automation stack is on the homepage.

One honest limit on that last step. SchedPilot’s upload call takes an image from a public URL, so your cards and slides need somewhere to live before anything can queue them. It is a ten-minute job once, and the step people forget when planning backwards from the calendar.

So the division is stable: a person reads once and marks, a machine does the rest. Run the pass on your best-performing article this week. If it yields fewer than twenty units, the article is a summary of something, which is a more useful finding than a month of posts.