Content for social platforms
Social content is better pulled out of one strong piece than built from scratch for every platform. The bottleneck is not production; it is having something worth saying, and that thing stays one thing even when its shape becomes four.
- Lesson 4 of 8
- Beginner
- Free, no signup
One piece, and the shape it takes in each place
The first column is the source and the next three are destinations. All of them come out of one point.
-
Your own site
2- The full piece
Where everything is written and can be found later.
- The sharing tags
Exactly what shows up in a messaging app preview.
- The full piece
-
Instagram
2- An image with six words
The image is read first, the caption second.
- An answer to a recurring question
The same question that arrives in direct messages.
- An image with six words
-
Telegram
2- Long text, uncut
A channel shows the text in full and it stays searchable.
- A file or a price list
Something meant to be found again.
- Long text, uncut
-
LinkedIn
2- A short story from a project
With what you learned, not with praise for yourself.
- One technical decision and its reason
For an audience that does the same work themselves.
- A short story from a project
These columns are not the stages of a workflow and their order does not matter. There is no number about reach on these platforms in this figure either, because we have not measured it.
Last checked: Facts and tool names in this lesson are re-checked against their sources on this date.
Do not build from scratch; reshape one piece
Small teams usually lose in the same place: they build content separately for every platform. The first week works, by the third the quality drops, and by the sixth everything has turned into quotes and stock photos.
The other route is repurposing: you build one complete, correct piece and then shape it for each place. That is not the same as copying one text into four places. Shaping means the same point is rewritten in that platform own format; copying means a text written for somewhere else sits where it is of no use.
The reason is simple. The hardest part of this work is not writing, it is having a point: an experience, a number you measured yourself, a mistake you made. That point does not run out from being told four times, but if you try to say something new twelve times a week it does run out, and then you are obliged to say something you do not have.
There is a real exception and it should be said: accounts where the platform is the product. Somebody whose work is making short video builds for that format from the start, and for them repurposing is a step backwards. This lesson is written for a business whose main place is its own site, with social as the route to it.
Which piece should you repurpose? The one that answered a real question and had somebody ask the next question afterwards. A piece nobody asked anything about usually has nothing to say elsewhere either. Newness is a poor criterion too; a piece written six months ago that is still correct is the best possible source.

A lot goes in, one thing survives, several shapes come out
The bottleneck is the middle, not the bottom. If the middle is empty, the bottom has nothing to say either.
What you already have
- The questions customers ask
- The work you did this month
- A mistake you made and fixed
- A source you read yourself
What comes out
- A long text for a channel
- An image with one sentence
- A short answer to a recurring question
- A short story from the same work
The figure shows the order of the work, not its yield. How many posts come out of one piece depends on the piece and has no fixed number.
What a native format means
A native format is something that is complete where it sits: you do not have to go anywhere else to understand it. A teaser that only says "link in bio" is not a native format; it is a promise whose cost is paid by the reader.
Take that seriously, because moving an audience off a platform always has a cost and that cost has to buy something. If the whole point fits in three lines, write the three lines and save the link for a piece that genuinely continues. A post that is complete and then says "the fuller version with the price table is here" is honest; a post left deliberately incomplete to earn a click works once.
A practical rule for choosing the format: every post should ask the reader for one specific action, and that action has to be possible on that platform. Saving it, sending it to one other person, answering a question, or going to a page. If you cannot say what exactly should happen after somebody sees this post, the post is not ready.
And something deliberately absent from this lesson: numbers. Caption length, the best hour to publish, how many hashtags, image sizes and algorithm behaviour all change every few months. Any number we wrote here could be wrong by the time you read it. The right reference for those numbers is the platform own help pages, and not any article in Persian, including this one.
There is one practical decision that changes everything on Instagram: the sentence that sits on the image matters more than the whole caption, because it is the first and sometimes the only thing read. That sentence has to mean something on its own, rather than being the title of something else. If you put a generic phrase like "important tips" on the image, you lost the reader right there.
Which platform carries the weight in Iran
What we see in daily work with clients is this: in Iran the weight sits on Instagram and Telegram, and the other platforms play a side role. That is an observation, not a statistic; we have not measured social platform share in Iran, and we do not write down a number we did not measure.
What is structural, though, and does not change, is worth saying. Telegram has channels, shows long text without cutting it, keeps files and has search inside itself, so it suits guides, price lists, files and anything meant to be found again later. Instagram is visual and the first thing read is the image, not the caption, so a point that exists only in the text goes half unread there. LinkedIn has a professional register and a working audience, and it is worth the effort for a business whose customer is another business.
A practical note for choosing: instead of asking which platform is bigger, ask where your customer asks questions. We have seen many teams producing on Instagram while all their real questions arrive in direct messages and on Telegram; answering exactly those questions is then the best content that can be made, because you already know somebody wants it.
And a warning: access and restrictions in Iran change and move the platform mix around. That is why building everything on one platform is a risk, and it is why the main piece stays on your own site.
There is a common mistake in this same choice: opening an account where your audience is not, only so as to be there too. Every extra platform is another content queue that has to be fed, and an account with no post for three months is worse than not being there, because it tells whoever finds it that the business has stopped.
One voice, several formats
Repurposing breaks in one specific place: when each platform finds its own voice. The site speaks formally, Instagram gets jokey, Telegram goes dry, and LinkedIn uses words nobody in the company says out loud. To somebody who follows you in two places, that reads as three different brands.
The simple test: put an Instagram caption and a paragraph from your site side by side and ask somebody outside the team whether one person wrote them or two. If the answer is two, the problem is not the format, it is the voice.
What helps is a one page note where you have written down: do we address the reader formally or casually, which words do we never use, and what is our tone in the face of a hard question. That note is later the input to the recipe below, and without it any language model applies its own default voice, which is usually the voice of a brochure.
The format, though, is free, and it should be. One point can be a seven line text on Telegram, an image with six words on Instagram, and a short story from a project on LinkedIn. Those are three translations of one sentence, not three personalities.
What the platform does not give you
The platform brings an audience and does not give you three things: ownership of that audience, a searchable archive, and control over who your post is shown to. All three stay in its hands and can change tomorrow.
The practical consequence is not to abandon social. It is to keep the main piece somewhere else. The full article lives on your own site, the posts point at it, and if one platform were gone tomorrow, what you built is still standing. The difference between the two situations becomes clear the day an account is lost, and by then it is too late to make up for it.
There is one more thing that gets less thought: when you send a page link in a messaging app, the preview the other person sees comes from the page itself, not from the app. The title, the description and the image beside the link are exactly what the page wrote in its sharing tags. So right where you thought the social content ended, there is one more piece of text still to be written, and most sites forget it.
There is a habit here that costs nothing: before publishing any post, know where the full version of that point lives. If the answer is "nowhere", then either the point is small enough that the post is all of it, or there is a piece you have not written yet and this post is standing in its place.
What the platform gives and what it keeps
This balance is not level and is not meant to be. It only says which side is in your hands.
What it gives
- An audience that is already there
- Fast spread for whatever catches on
- Direct conversation with people
What it keeps
- The link to the audience, which belongs to the account
- The archive, which is not findable outside
- The decision about who sees the post
The conclusion is not to leave the platform. It is not to keep there the thing whose loss would be expensive for you.
The fast path, with AI
The usual move is to ask a model to "write a few posts for this article" and get back something colourful with nothing of the article in it. The professional version adds two constraints that make all the difference: the model may not say anything that is not in the source text, and if the text has nothing for a platform it has to say so rather than fill the gap.
- Pick a piece that genuinely has something in it, not the newest one. A piece that is empty in itself will not yield good posts from any recipe.
- Prepare the voice note and one sentence per platform: exactly what action you want from the audience in each place. The model does not know this, and if you do not write it, it decides for you.
- A cheap, fast model is enough for this pass, because the work is selecting and reshaping rather than judging. The current ranking is listed in this site AI section.
- Run the recipe below. Rule six is the important one: the model has to say which part of the source each post came from, so that checking takes seconds rather than minutes.
- Read the outputs together, not one at a time. That is the only place where a break in the voice shows up, and fixing it is a short manual pass.
Copy-ready recipe
Source text:
{the whole article}
Our voice: {formal or casual, the words we do not use}
Reader: {the reader in one sentence}
The action we want on each platform: {Telegram: ...} {Instagram: ...} {LinkedIn: ...}
1) Write one version for each platform above, in that platform native format.
2) Use only the points and facts in this text. Add no new point.
3) Write no number, date, percentage or name that is not in the text.
4) If the text has nothing for a platform, instead of writing a post write "this text has nothing for this platform" and give the reason in one sentence.
5) Each version must be complete in itself and ask for the one action requested. Do not ask for more than one.
6) Under each version, write which paragraph of the source it came from.
7) Do not suggest hashtags and do not claim that anything is trending.
Before you trust the output: Check three things yourself before publishing. First, every number and name left in the output must be in the source text; rule three exists exactly for that and it does get broken. Second, the voice: read the outputs one after another and see whether one person wrote them or three. Third, the platform own rules, from image sizes to advertising policy, which the model does not have an up to date copy of and whose reference is that platform help pages.
AI in this kind of work
Our position is plain: use AI to shape something you have, not to manufacture something you do not. Social is where that line gets crossed more than anywhere else, because the pressure to publish daily is there and a model produces text without tiring. The result is accounts that post every day and that nobody remembers anything from.
Tools that actually help
- Claude It holds a negative constraint across a multi part output, and the whole value of the recipe above sits in those constraints: say nothing outside the text, write no number that is not there. Iran is on neither of Anthropic two supported country lists; we read that on Anthropic own page.
- Gemini It understands Persian itself, so it both produces Persian output and can be asked which version drifted away from the voice note. Google own page says the app works in over 230 countries and territories and more than 70 languages, and Iran is not on that list.
- NotebookLM It earns its place when there are several sources, because it works only on the files and links you gave it, which enforces the "say nothing outside the text" constraint at the tool level. Google own help says it runs in the same regions where the Gemini app runs, and Iran is not on that list.
Where it backfires
Two risks, both specific to this job. First, a model speaks confidently about the platforms themselves: what is trending, how many hashtags, which hour is better. Those either come from stale data or from nowhere at all, and unlike a technical claim nobody checks them. The only reference for this is the platform own help. Second, images: Google image models mark their output with SynthID, an imperceptible watermark Google considers detectable, and platforms have added their own labelling of AI content. So an account publishing generated images every day is identifiable. The lesson on AI image watermarks covers this in full.
Sources: Google DeepMind: SynthID Google Search Central: Google Search and AI-generated content Anthropic: supported countries and regions Google: where Gemini Apps are available
Where this advice stops
There is no platform number in this lesson and that is a choice rather than a gap: caption length, publishing hour, hashtag counts and algorithm behaviour change every few months and their reference is the platform own help. Two other boundaries. First, repurposing works when the source piece has something in it; no recipe pulls good posts out of an empty text. Second, this lesson is written for a business whose main place is its own site; if the platform itself is your product, the order reverses and you build for that format from the start.
From our own work
Something we ran into on this site ourselves, which shows that a link preview is the page job and not the messaging app. The private proposals section of this site is hidden from search engines by five layers: an X-Robots-Tag header, a robots meta tag, a robots.txt disallow for all four languages, no public single URL, and page caching switched off. With all of that, the OpenGraph tags are deliberately rendered on the lock screen, and the theme code says why: messaging apps read only OG and pay no attention to robots. The result is that the very page Google is not allowed to index builds a correct, complete preview when its link is sent on WhatsApp. The lesson for you: what appears beside your link in a messaging app is text you wrote on the page yourself, and if you do not write it, the app makes something up.
Real follow-up questions
How many posts a day should I publish?
A number you can still keep up next month, and that number is set by your capacity, not by the platform. A team publishing two good posts a week is ahead of one that posts three a day for ten days and then disappears for three weeks. Deciding the rhythm is the subject of the next lesson in this path.
Should the caption be short or long?
It depends on what the caption is supposed to do. If the image already made the point, the caption only needs to say what to do next, and it is short. If the point is in the text, the caption is the piece and has to be complete, but then you should ask why it is not on the image, because the image is read first.
Is there a problem with posting the site article as it is on Telegram?
Not for SEO, because a social channel does not compete with your site page for ranking. For reading it usually does not work, though: a text written for a web page with headings and an index becomes heavy inside one continuous message. Better to write the same point again for that format and put the link to the full piece under it.