How to Post on LinkedIn Consistently (Without It Eating Your Week)

You already know how to post on LinkedIn consistently, in theory: pick a rhythm and hold it. What actually happens is that week three arrives, the good week is over, and you are looking at an empty composer with nothing in your head you want to say out loud. That is not a discipline failure, and treating it as one is exactly why the next attempt fails the same way — it is a supply failure, and willpower does not manufacture supply. This post is about the pipeline that keeps the queue full: where the raw material genuinely comes from, which parts of the work a machine should touch, what the whole thing honestly costs, and how to tell whether it is working.

Consistency is a supply problem

Ask a founder why they stopped posting and you will almost always get a version of "I got busy". Push once and something more specific comes out: they had four or five things genuinely worth saying, they said them, and the well was dry. The first fortnight felt easy because it was spending a backlog that had accumulated over years of doing the work. Nothing replaced it, because nothing in the process was designed to replace it. What looks like a motivation problem is an inventory problem wearing a motivation problem's clothes.

The distinction matters because the two problems have opposite fixes. If consistency were about discipline, the answer would be accountability: a partner, a streak, a public commitment, a reminder at nine on Tuesday morning. All of those work for about a fortnight, which is precisely long enough to convince you they are working. If consistency is about supply, none of them do anything at all, because a reminder to post is not a thing to post about. You end up publishing something you do not really believe in order to protect a streak, and that costs more credibility than the missed day would have.

You can test which one you have in about a minute. Open a blank document and try to write down ten things you would be happy to defend in public — not headlines, actual positions, each with a reason attached. If you reach ten without straining, you have a discipline problem and you should stop reading and go and post. Most people we run this with get to four. The gap between four and ten is the entire problem, and it does not close by trying harder.

There is a second-order effect that makes the dry spell worse than it first looks. Once the backlog is spent, every post has to be invented from nothing at the moment of posting, which is the most expensive possible way to produce one. It is also the moment you have the least perspective, because you are writing to fill a slot rather than because something is true. That is how a feed slides from things a person actually thinks into observations nobody could possibly argue with, and a post nobody would argue with is a post nobody remembers.

Why a content calendar does not fix it

The standard response to falling off is to get organised: a spreadsheet with dates down one side and themes across the top, or a scheduling tool with a queue you fill in advance. It feels like progress because it produces an artefact, and the artefact looks like a system. It is not one. A calendar is a commitment device — it records what you intend to publish and when — and it has no opinion whatsoever about where the material is going to come from.

This is why the three usual purchases disappoint. Each one is a real fix for a real problem, and none of them is a fix for the problem most people have when they buy it. The batch day genuinely reduces context-switching, the scheduler genuinely publishes on time, and a ghostwriter genuinely writes better sentences than you do on a Thursday evening. What none of them does is generate the raw material, which means all three can work perfectly while your output still stops in week three.

Notice that the third row is the expensive one, and it is expensive in a way that is easy to miss. A ghostwriter needs input, so you end up in a weekly call where someone extracts opinions from you — which is the capture step, done manually, by a person you are paying. That can be a perfectly good arrangement. It just means the thing you actually bought was capture, and you could have built that part yourself for the cost of a shared note and a recurring calendar block.

Where LinkedIn posts actually come from

If the queue is the problem, the queue is where the work belongs — and the first thing to accept is that you are not going to have ten new ideas a week. Nobody does, including the people whose feeds look like they do. What you do have is a week of work in which you explain things, decide things, disagree with things and fix things, and almost all of it is post-shaped before you touch it. The raw material is not missing. It is simply not being captured, so it evaporates within a day or two of happening.

The reason these feel like nothing is that they happened to you. Expertise makes its own knowledge invisible: the answer you have given three times this month is boring to you and genuinely useful to everyone who has not spent ten years in your industry. This is the single most reliable pattern we see. When a founder says they have nothing to say, and you sit and read their last two weeks of sent email with them, the problem inverts within twenty minutes — the list is too long rather than too short.

The practical form of this is a capture habit, not a content strategy. One place — a note on your phone, a channel you message, a shared document — and one rule: when you finish explaining something, write down what you just explained, in the words you just used. Do not write a post. Write the argument in two sentences and move on. The whole thing takes fifteen seconds and it is the step that everything downstream depends on, which is why it is also the step that quietly does not happen.

What you are building here is an inventory, and inventories have a property that makes them worth the trouble: they compound. A note captured in March is still usable in July, and a good argument can be told from three different angles across a year without repeating itself. That is the same mechanism we describe in content multiplication — one piece of thinking, several finished outputs — applied at the level of the raw idea rather than the finished asset.

The pipeline: capture, shape, queue

Once material is being captured, the rest is a small amount of plumbing. The shape that works is three stages that run on different clocks: capture runs continuously and costs seconds, shaping runs once a week and costs about an hour, and publishing runs on its own schedule and costs nothing because the work is already done. Keeping those clocks separate is the whole trick, because the reason a Tuesday post is hard is that people try to do all three stages at 8:45 on Tuesday morning.

The tooling here matters less than the shape, and it is worth being blunt about that because the tool is the fun part and the shape is the part that works. We build these in n8n when the workflow needs branching, retries and a place to store state between runs, and in Zapier or Make.com when it is genuinely linear and the client already lives in one of them. All three will move a transcript into a document and a document into a queue. Choosing between them is an hour of thought, not a week, and it is reversible.

The one stage that is not negotiable is the review gate. Nothing publishes without a person reading it, and the person is you until the voice is stable enough to hand over. This is not caution for its own sake — it is the only thing standing between a working pipeline and an account that quietly starts sounding like everyone else's. A queue that publishes unread is not a content system, it is a liability with a schedule attached.

A worked example

Here is what this looks like as actual files, because the abstraction hides how ordinary it is. We build the shaping half in Claude Code against a folder in the client's own repository or drive, so the rules that govern the writing are text files the client can read and change rather than a prompt buried in a tool nobody can open. The folder below is roughly what a working setup contains after a month.

The weekly run is one command and about forty minutes of human attention. Claude Code reads the week's capture file, groups the notes by argument rather than by topic, and drafts three angles for each cluster it thinks will hold — the objection version, the worked-example version, and the one-sentence version. It writes them into queue/ as separate files with the source note quoted at the top, so you can always see which of your own sentences a draft came from.

Then a person reads them, which is the part that does not compress. Roughly half get deleted immediately, because a draft that argues something you do not quite believe is obvious within one sentence. Of the rest, most need one real edit: a specific detail put back in, a hedge taken out, or an ending that stops one sentence earlier. Forty minutes on a Friday reliably produces the following week and usually some of the week after, which is what "the queue is full" actually means in practice.

What to automate, and what never to

The failure mode on the other side of this is over-automation, and it is worse than the empty calendar because it is harder to see. An account posting three times a week of competent, structured, entirely forgettable content looks like it is working from the outside, and it is producing nothing — no replies worth having, no conversations, no reason for anyone to remember who you are. The way to avoid it is to be precise about which part of the work is mechanical and which part is the actual product.

The practical rule we hold to is that a machine may move, reshape and format your thinking, and may never originate it. A model asked to generate ten LinkedIn post ideas for a consultancy will produce ten plausible ideas, none of which you believe more than anyone else does, and that is the exact texture readers have learned to scroll past. The same model asked to turn a note that starts "the thing clients get wrong about handover is" into three drafts is doing genuinely useful work, because the belief was already there and only the shaping was missing.

There is a second rule that costs people more than the first: do not automate the reply. Comments and DMs are where a post turns into a conversation and a conversation turns into a call, and they are the highest-leverage twenty minutes in the whole process. Automating them saves the cheapest part of the work and destroys the part that pays. If the pipeline is working, you should have more time for replies than you did before, not less.

Keeping it sounding like you

Voice drift is gradual and it is measured in months rather than posts, which is why it is rarely caught in time. Each individual draft is a small compromise — a slightly more general opening, a hedge left in, an ending softened — and none of them is worth arguing with on its own. Six months later the feed reads like a competent stranger. The defence is a written standard and a check that runs before publishing, not a feeling you consult when you remember to.

The most useful of these is the disagreement test, and it is the one people quietly skip. If nobody could reasonably argue with a post, it carries no information — it is the content equivalent of saying that quality matters. This does not mean manufacturing controversy, which reads as exactly what it is. It means being willing to say the specific version of the thing: not "documentation is important", but "most process documentation fails because it was written by the person who already knows the process".

Keeping a written voice.md makes this enforceable rather than aspirational, and it has an obvious second benefit: it is the artefact you hand to a new team member, a contractor, or a model, and all three then produce closer drafts on the first attempt. That is the same principle that makes the rest of an operating system work — the standard lives outside your head, so it survives you being busy. We go further into how that gets structured in content strategy for founder-led businesses.

What it actually costs

The honest version is that this is worse than doing it by hand for about a month, and the reason is structural rather than a matter of skill. You are spending the first fortnight capturing rather than publishing, which feels like going backwards, and the first drafts a system produces are always worse than the ones you write when you are fired up about something. People abandon this in week three, and week three is exactly where the compounding starts.

There is also a cost nobody advertises: someone still has to have opinions. This system makes it cheap to publish what you think and does nothing whatsoever to make you think it. If a founder genuinely has no distinct position on their own field — which is rare, but it happens, usually in businesses that compete on price — then no pipeline will help, and the honest advice is to spend the effort somewhere else. Every piece of this depends on there being something in the capture step worth shaping.

The other real cost is the maintenance one, and it is small but not zero. Roughly once a quarter the voice file needs updating because your positions have moved, and the clustering step needs looking at because it starts grouping by topic rather than by argument as the archive grows. Budget an hour a quarter. A system nobody maintains degrades quietly, which is a pattern that shows up everywhere and is the reason templates fail serious businesses — the structure was correct on the day it was made and nothing kept it true.

Knowing whether it is working

Impressions are the metric LinkedIn puts in front of you and close to the least useful one available, because they measure distribution rather than effect. A post can reach forty thousand people and produce nothing, and a post can reach eleven hundred and produce two conversations that turn into work. What you actually want to know is whether the feed is moving people along a path, and that path has rungs you can watch in order.

Read those in order when something is not working, because each rung diagnoses a different failure. Stuck at rung one is a supply problem and everything above this section applies. Stuck at rung two means you are reaching people but giving them no reason to come back, which is usually a topic that wanders. Stuck at rung three — reach and no substantive replies — is nearly always the specificity problem: the posts are correct and general, and nobody argues with correct and general.

The number worth tracking is the one nearly nobody tracks: how many inbound conversations mentioned something you wrote. It requires you to ask, and it takes a line in whatever you already use to record leads. It is the only figure in this whole exercise that connects the work to revenue, and it is far more stable month to month than impressions, which swing on things you do not control. If it is above zero and rising slowly, the system is working, whatever the reach chart is doing.

Frequently asked questions

How often should I actually post on LinkedIn?

Three times a week is enough for almost every founder-led business, and it beats five times a week for a month followed by silence. Frequency only compounds if it is survivable: the rhythm your audience learns to expect is the one you can still hold in a bad week. Pick the number you could hit during your busiest fortnight of the year, and treat anything above it as a bonus rather than the baseline.

Is it better to post consistently or to post well?

Consistently, but only because consistency is what makes posting well possible. One good post in a quiet month has nothing around it and nothing to build on, while a steady rhythm lets every post inherit the credibility of the ones before it. The trap is reading that as permission to publish filler, and a post nobody would disagree with costs you more attention than a missed day ever does.

Should I use AI to write my LinkedIn posts?

Use it to shape material you have already produced, not to invent opinions you do not hold. A model is genuinely good at turning a rambling voice note into a structured draft, grouping a week of notes by argument, and cutting a draft by a third without losing the point. It is bad at knowing which of your beliefs is worth defending, and a feed written entirely by a model reads as one, because every post arrives equally confident and equally unremarkable.

What do I post about when I have run out of ideas?

You have not run out of ideas, you have run out of captured ones. Look at the last two weeks of your actual work — client questions, sales objections, decisions you made, things you fixed, approaches you rejected — and write down what happened instead of trying to think of a topic. Almost everything on that list is already post-shaped, and the only reason it feels like nothing is that it happened to you and therefore seems obvious.

Do scheduling tools solve consistency?

No. They solve timing, which is a different problem. A scheduler reliably publishes whatever is in the queue at a sensible hour, and it has no opinion at all about whether the queue is empty. Buying one while the queue is the problem is the most common wasted purchase in this category, because the tool works exactly as advertised and the outcome does not change.

How long before a content pipeline actually produces anything?

Expect two to three weeks before the queue holds real stock, and around two months before the output beats what you would have written by hand. The first fortnight goes on capturing rather than publishing, which feels like going backwards. The honest version is that month one is usually worse than doing it manually, and the payoff is that month six exists at all.

Can I build this myself, or do I need someone to build it?

You can build the first version yourself in a weekend, and most people should. A shared note for capture, a recurring calendar block for shaping, and one document holding the queue will take you further than most tooling will. Outside help earns its place at the point where the workflow is right but nobody has time to run it, which is a different problem from not knowing what to build.

Where to start

Do not build the pipeline this week. Open a note on your phone, call it something you will recognise, and for the next ten working days write down every question anyone asks you that takes more than a sentence to answer — the question in their words, your answer in yours, thirty seconds a time. That is the capture step, done by hand, and it is the only part of this that cannot be skipped or bought. At the end of the ten days you will have between fifteen and thirty entries and a much better idea of whether you have a supply problem or a discipline one.

Then automate exactly one stage: whichever one you noticed yourself avoiding. For most people that is the shaping hour, which is where a tool genuinely earns its place. If you would rather the whole thing existed already, that is what we build — the capture route, the shaping rules, the review gate and the queue, handed over as files your team runs without us. It is described in full on our LinkedIn content system page.

Keep reading

• Content Multiplication — once the queue holds, the same thinking should be reaching more than one platform.
• Content Strategy for Founder-Led Businesses — the layer above this one: what the whole feed is for.
• How to Create Content That Compounds Over Time — why an inventory of arguments is worth more than a run of posts.