The symptom is easy to recognise: you publish the same idea on four networks, and a follower who uses three of them scrolls past your post three times in an afternoon without stopping on any of them. The content was fine. The problem is that it was one post wearing four different hats, and everyone could tell.

Writing variations is a separate skill from writing posts, and it is the skill that decides whether repurposing compounds reach or dilutes it. The good news is that it is mostly mechanical once you know which levers to pull. This guide names the levers, gives you a rewrite method that takes about ninety seconds per destination, and lists the traps that make a variation read as a copy.

Why identical captions cost you twice

Two different failures hide behind one symptom, and they need different fixes.

Your audience sees the duplicate. The overlap between your platforms is larger than it feels. Followers who care about you follow you everywhere, which means the people most likely to see your post twice are the ones you most want to impress. The second sighting does not just fail to land — it spends some of the goodwill the first one built.

The platform sees the duplicate. Networks detect the signals of low-effort distribution: identical caption text, media already published elsewhere, visible watermarks from a competing app. You will not receive an error. Your reach simply sits below what the same content would have earned natively, and you will never know which post it happened to.

The fix in both cases is the same: make the text genuinely different, not cosmetically different.

The five levers of a real variation

A variation is real when at least three of these five change. If only one changes, you have edited a caption; if none change, you have copy-pasted.

• The first line. This is the whole ball game on platforms where the opening line is what the feed shows. Rewrite it from scratch for each destination rather than trimming the original.
• The shape. A paragraph on LinkedIn, a thread-ready block on X, a conversational question on Threads, a caption under a visual on Instagram. Same idea, different container.
• The length. Bluesky gives you 300 characters, X 280, Threads 500, LinkedIn 3,000, Instagram 2,200, Tumblr 4,096. Length is not a constraint to survive; it is a lever. A 280-character version forces you to find the sharpest sentence in the idea, and that sentence is often the better hook everywhere.
• The call to action. "Read the full breakdown" and "Tell me which of these you have tried" are different posts with different jobs. Match the CTA to what the platform is for — conversation, click, save, or share.
• The evidence. Lead with the number on one platform, the story on another, the screenshot on a third. The claim stays constant; the proof changes shape.

Change three, keep the claim. That is the entire discipline.

A four-pass rewrite that takes ninety seconds

Do the passes in order. Skipping the first pass is the most common reason a variation still reads as a copy.

• Pass one — restate the claim in one sentence, without your original wording. If you cannot, you do not yet know what the post is about, and no amount of trimming will rescue it.
• Pass two — write a new first line for this destination. Different audience state, different opening. Someone arriving from a feed that rewards conversational openers should not get a line written for a feed that rewards declarative authority.
• Pass three — cut or expand to the destination's length. Work with the character counter rather than against it. When the counter forces a cut, cut the second-best example, never the claim.
• Pass four — swap the CTA and the media treatment. One platform wants the link, one wants the question, one wants nothing at all. One wants a 4:5 crop, one a 9:16 frame.

Ninety seconds is not a target for its own sake. It is the point at which rewriting stops feeling like overhead and starts feeling like editing, which is what keeps people doing it on week six.

What to change and what to hold constant

The distinction matters more than the amount of change.

Hold constant: the underlying claim, the brand voice, the facts, the destination URL and its tracking. If the claim shifts between platforms, you do not have variations — you have four posts that contradict each other, and someone will screenshot them side by side.

Change: the opening, the structure, the length, the CTA, the crop, the hashtags or tags where the platform actually uses them, and the posting time if the audiences are in different timezones.

A useful test before you publish: read the two versions back to back and ask whether a person who follows you on both would recognise them as the same idea without feeling they were reading the same post twice. If they would not recognise the connection, you have rewritten too far. If they would feel the repetition, you have rewritten too little.

The traps that make variations read as copies

• Changing only the hashtags. The body is identical and everyone can see it. Tags are a suffix, not a variation.
• Changing one word in the first line. "Here is how we…" becoming "Here's how we…" is not a new opening.
• Syndicating the same link with the same preview across every network. The card does the recognising for the audience before your text is even read.
• Watermarked video moved between apps. Re-upload from a clean master; a rival network's logo in the corner is a demotion signal on the destination.
• Writing the longest version first and only ever cutting. You end up with five abridged versions of the LinkedIn post. Write the version each platform wants, even if that means writing the short one first.
• Keeping the same crop everywhere. A 1:1 image dropped into a 9:16 slot gets cropped at the worst possible moment, usually across a face or a chart label.

Where AI helps, and where it does not

An assistant is genuinely useful for the middle of this job: given a finished caption and a destination, it will produce three structurally different drafts faster than you can. That is the mechanical part, and automating it is sensible.

What it does not do reliably is know which of your sentences is the load-bearing one. Variations that fail usually fail because the rewrite preserved the filler and lost the claim. Keep a one-page voice brief and a small bank of captions that already worked on each platform, and feed both in; a model given your winners produces variations in your voice rather than in generic marketer voice. The AI captions guide covers that setup in detail.

The rule of thumb: automate the restructure, never the hook and never the claim.

Where PostHub fits, and where it does not

PostHub removes the mechanical friction around variations. You write one draft, then add per-platform overrides for the destinations that need their own text, so the LinkedIn version and the X version live on the same post instead of in two documents. The composer shows the tightest character limit across your selected accounts in the global counter, then switches to each platform's own limit inside its override box — which means a 3,000-character LinkedIn post and a 280-character X post coexist without you arithmetic-ing the difference. The platform preview shows how the post will actually render on each destination before it goes out.

Where it does not help: deciding what to change. No tool can tell you that your opening line is the part that needs rewriting, or that this particular idea deserves a question on Threads and a number on LinkedIn. Variations are a judgement problem wrapped in a mechanical problem — automate the mechanics, and keep the judgement.

Common variation mistakes and fixes

• Mistake: one caption, four networks, hashtags swapped. Fix: change the first line, the shape and the CTA — three of the five levers, minimum.
• Mistake: writing the longest version and only ever cutting. Fix: write the version each platform wants, short first if the platform is short.
• Mistake: the claim drifts between platforms. Fix: state the claim in one sentence before you start, and keep that sentence true everywhere.
• Mistake: variations take an hour each. Fix: run the four-pass method on a timer until it takes ninety seconds, then keep the timer.
• Mistake: every variation carries the same link preview. Fix: attach the link where it does not cost distribution, and let the text stand alone where it does.

Key takeaways

• A variation is real when at least three of five levers change: first line, shape, length, call to action, evidence.
• Hold the claim, the voice, the facts and the tracking constant; change everything else.
• Rewrite the first line from scratch for every destination — it is the single change with the highest return.
• Use the platform's character limit as a creative constraint, not an obstacle to survive.
• Automate the restructure with AI; never hand it the hook or the claim.

FAQ

How different does a variation have to be?

Different enough that the opening line, the structure and the call to action are not recognisably the same, and not so different that the underlying claim changes. Three of the five levers is a good working threshold; two usually reads as a copy.

Is it worth writing variations for platforms I post to twice a week?

Yes, and the reason is not fairness to the platform. A variation takes under two minutes once the method is habit, and the alternative — an identical caption on a low-frequency platform — is the most visible kind of repetition, because your followers only see you there occasionally and the duplicate stands out.

Should I vary the media too?

Where the formats differ, yes. The same claim can be a chart on one network and a photo on another, and the crop should match the destination's aspect ratio rather than being letterboxed into it. The media requirements guide lists what each network accepts; the cross-posting strategy covers the wider adaptation rules for video and carousels.