I love magic tricks.
I can still remember the sense of wonder I had as a child watching a magician make something disappear. There was something almost magical about the idea that somebody could take an ordinary coin, card or object, do something I couldn’t quite follow and then make it vanish in front of my eyes. I knew, even then, that there had to be an explanation, but that was part of the fascination. The magician wasn’t really making anything disappear. They were making me look in the wrong place. Somehow, while I was concentrating on one hand, the important part of the trick was happening somewhere else, and by the time I realised that, it was too late. The best magic left me with that slightly wonderful feeling of knowing I had been fooled without being able to work out how.
The phrase “Now you see me, now you don't" or “Now you see it, now you don’t” has become associated with people or things appearing and disappearing, particularly in the language of magic and illusion. As far as I can tell, there isn’t a reliably documented magician who can be credited with inventing the saying, which is perhaps appropriate. It seems right that a phrase about things mysteriously appearing and disappearing ought to have a mysterious history. Despite this, a movie franchise has been established on the saying with the films. Now You See Me appeared in 2013, followed by Now You See Me 2 in 2016 and Now You See Me: Now You Don’t in 2025. The connection with magic is explicit in those films; the central characters are illusionists, and the boundary between performing a magic trick and carrying out an elaborate deception is blurred.
It’s also had a life beyond the magician’s stage. Isaac Asimov played with almost exactly the same idea in Second Foundation, whose two original sections were published as Now You See It... and ...And Now You Don’t. In Asimov’s case, there was considerably more going on than somebody making a coin disappear, but the idea is the same. Something is there; you think you know where it is, and then suddenly you’re not quite sure.
Which brings me, in a roundabout way, to LinkedIn. Because posting on LinkedIn and being seen seems like a magic trick. And I think I've already mentioned that I love magic tricks ... but I don't love this one
posting on LinkedIn and being seen seems like a magic trick. And I think I've already mentioned that I love magic tricks ... but I don't love this one
You see, I’ve been trying to work out how LinkedIn decides whether anyone is going to see one of my articles. Some people see me...some people don't. It’s not that I have no idea whether anybody reads them. In fact, LinkedIn gives me quite a lot of information about that. Articles have figures for impressions, members reached and views, so I can see how many people have been given the opportunity to see an article and how many have actually looked at it.
In my case, the proportion of views to members reached is often reasonably high, usually somewhere between 30 and 50 per cent. So when LinkedIn puts an article in front of somebody, there’s a fairly good chance that person will actually look at it. The curious part of the trick obviously happens before the audience arrives.
Some of my articles get thousands of impressions almost immediately. And yet others can sit there for a day or more with fewer than 100. Now I’m obviously not expecting every article to get thousands of impressions. Some of the subjects I write about are quite specialised, and I’m perfectly realistic about the size of the potential audience. There probably aren’t thousands of people on LinkedIn sitting around waiting for my latest thoughts on information standards, analysis / statistical methods or NHS data. But there are a few, and I’m very grateful to them. What I find difficult to understand, though, is why articles that appear broadly comparable to me can receive such dramatically different treatment. One can be given thousands of impressions almost immediately (now they see me), while another can appear to have been put into a small box, the lid closed and the box moved backstage (now they don't). It hasn’t actually disappeared, of course. If I go to my own articles, it's still sitting there quite happily. It simply hasn’t been shown to very many people.
What I find difficult to understand, though, is why articles that appear broadly comparable to me can receive such dramatically different treatment.
I can accept that subject, title and timing will make a difference, although I usually post at the same time of the day. An article that happens to connect with something people are already talking about may attract more attention than something I’ve been thinking about for three months but which nobody else has noticed. Some subjects are inevitably going to have a larger potential audience than others, and some titles will work better than others. What I’m trying to understand is why the variation can be so large when the articles themselves seem reasonably similar.
The obvious explanation is the algorithm, which is both perfectly plausible and entirely unsatisfying. The algorithm decided to distribute one article widely and another one narrowly. Fair enough. But that’s a little like explaining a magic trick by saying that the magician did it with magic. It describes the outcome without telling me very much about what happened. Was it the subject, the title, the timing, the people who happened to be online, the response to the first few impressions, the relationship between my article and the interests LinkedIn has assigned to particular members, or something else entirely? There may be a perfectly rational explanation inside the system, but from the outside it’s difficult to tell whether I’m looking at a sophisticated distribution mechanism or an elaborate sleight of hand.
And there certainly isn’t any shortage of people willing to explain how the trick is done. There’s a whole industry devoted to telling us how to get noticed on LinkedIn, with advice about hooks, posting times, engagement, comments, formatting, personal branding, storytelling, the number of lines before the “see more” button and the importance of getting people to interact with something as quickly as possible. Some of it is undoubtedly useful. Some of it is probably based on considerable experience. Some of it feels like being given instructions for performing a card trick by somebody who has spent several years studying card tricks but still refuses to tell you where to buy the cards.
There’s a whole industry devoted to telling us how to get noticed on LinkedIn, with advice about hooks, posting times, engagement, comments, formatting, personal branding, storytelling, the number of lines before the “see more” button and the importance of getting people to interact with something as quickly as possible
I’ve read some of the advice, and I’m sure some of it would improve my reach. I could make my articles shorter, make the openings more dramatic, choose subjects that are more likely to attract attention and turn more of my ideas into lists. I could make sure every article contains exactly five things leaders need to know, seven mistakes organisations are making or ten lessons nobody has told us about whatever the subject happens to be that week.
The problem with that approach, from my point of view, is that I don’t always have five things. Sometimes I’ve got three. Sometimes I’ve got one thing that takes 2,000 words to explain. I could produce another two, but that would start to feel like the literary equivalent of putting an extra card up your sleeve simply because the trick requires one.
I know the way that I write is probably part of the problem. I tend to start with something that has caught my attention and then see where it goes. Sometimes I know where I’m going, and sometimes I discover it along the way. I’m quite happy for an article to wander a little if that’s what the subject requires, because that’s how conversations work and it’s how most of us actually think.
We don’t necessarily start with a five-point framework, deliver point one, point two, and point three, and then arrive neatly at the conclusion. We start somewhere, remember something else, make a connection, change our minds a little and eventually work out what we’re trying to say.
I could, of course, make everything more LinkedIn friendly. I could open with some grand statement about what I learned from thirty years in the NHS, follow it with a story about a conversation I had with a chief executive in 2017, extract three lessons from the experience and finish by asking what other leaders have learned from their own journey. It would probably look very familiar, and it might even work.
But there’s a difference between learning how to communicate more effectively and learning how to perform a trick for the distribution system. I’m interested in the first. I’m less interested in becoming a professional magician whose principal audience is an algorithm.
there’s a difference between learning how to communicate more effectively and learning how to perform a trick for the distribution system. I’m interested in the first.
Then there’s AI, which introduces another possible piece of misdirection into the whole business.
You see, I do use AI when I write, and I’m quite happy to admit that. I use it to help structure my ideas, remove duplication, improve clarity (sometimes a lot) and occasionally to point out that I’ve managed to make essentially the same point several times without noticing. After several thousand words, it can be surprisingly difficult to see that the paragraph you’ve just spent half an hour rewriting is actually saying the same thing as one you wrote much earlier.
The important point for me, and hopefully for LinkedIn, is that all the ideas are mine, the experiences written about are mine, the examples I give are mine and the arguments I make are mine.
Just like in using AI, a magician can use a particular deck of cards, a table with a secret compartment and a carefully positioned mirror, but the equipment doesn’t explain the whole trick. You still need to know what the performer is doing. And that makes me wonder a little about how LinkedIn distinguishes between AI-generated content and something written by a person who has used AI as one of the tools in the writing process. I’ve got no evidence that LinkedIn is restricting my articles because I use AI, so I’m not suggesting that some secret system is examining my writing and deciding that this particular article deserves 63 impressions because it has detected suspicious levels of artificial assistance.
What I do find ironic, though, is that LinkedIn itself has introduced AI tools to help people write and rewrite their content. So we have a platform providing tools that can help people produce and modify writing while simultaneously having to deal with the enormous amount of generic AI-generated material now appearing on the platform. Introducing a “Seems like AI slop button” seems a bizarre thing to do when you have encouraged the practice in the first place.
we have a platform providing tools that can help people produce and modify writing while simultaneously having to deal with the enormous amount of generic AI-generated material now appearing on the platform. Introducing a “Seems like AI slop button” seems a bizarre thing to do when you have encouraged the practice in the first place.
Regardless of the reason, some material has become so recognisable as AI that it has almost developed its own magic trick. You get the dramatic opening, the short paragraphs, the list of lessons, the leadership revelation and the question at the end. The structure makes an appearance, performs its routine and disappears again, ready to reappear in somebody else’s post tomorrow.
I’m not saying that all AI-assisted writing looks like that, or that everything written in that style is AI generated. I don’t think it’s that simple. But it does raise an interesting question about how a platform might distinguish between something generated by a machine, something written by a person who happens to have used AI to tidy it up, and something written by someone who is an experienced writer and uses many of these techniques already. Maybe the Claude watermark will help in this – I’m sure other AI models have similar things within their text or will do soon.
In the meantime, if I write an article myself and then use AI to remove repetition, has the article become AI-generated? If I use it to restructure a couple of paragraphs, is that different? If I ask it whether an argument makes sense, where exactly does the magic boundary lie?
I don’t have an easy answer to that, and I’m not sure anybody else has a particularly simple one either.
All of this is important because I care about my reach on LinkedIn. I run a training and consultancy business, and I want people to know what I do. If somebody reads an article and then gets in touch because they think I might be able to help them, that’s obviously useful. I’m not writing several thousand words twice a week simply because I have an unusually strong attachment to the sound of my own keyboard. No, there’s a commercial reason for my writing, as well as the fact that I enjoy it and find that writing helps me think.
I’m not writing several thousand words twice a week simply because I have an unusually strong attachment to the sound of my own keyboard. No, there’s a commercial reason for my writing, as well as the fact that I enjoy it and find that writing helps me think.
What I don’t want to do is end up writing primarily for the algorithm. I could certainly change the way I write if that became my main objective, but I’m not convinced I’d still be writing about the things that actually interest me or are relevant to my business offering.
So I’ll carry on doing what I’ve been doing. I’ll continue to publish at least twice a week and write about the subjects that interest me, even when I know they’re unlikely to appeal to everybody. I’ll continue to use AI where it’s useful, particularly for structuring ideas, removing duplication and checking whether something I’ve written is actually clear.
I’ll keep an eye on the numbers because I’m interested in them, and because it would be rather foolish to ignore information that’s available to me. I’m just not going to pretend that I’ve discovered some secret formula for making LinkedIn distribute my articles. After all, I’ve spent much of my career telling people to start with the question, look carefully at the data and be cautious about explanations that aren’t supported by evidence. So I suppose I should follow my own advice. I can see what happened. I can measure it. I can compare it. I can ask questions about it. What I can’t yet do is explain it.
There's probably a perfectly logical explanation buried somewhere inside LinkedIn’s distribution system. There might be dozens of variables interacting in ways that would become obvious if I had access to the right data. I'm sure somebody at LinkedIn could explain it all to me in ten minutes, although I suspect that explanation might involve considerably more than the word “algorithm”.
I'm sure somebody at LinkedIn could explain it all to me in ten minutes, although I suspect that explanation might involve considerably more than the word “algorithm”.
That’s where the title of this article comes in. The article itself doesn’t vanish. It remains there, waiting for somebody to find it. What changes is whether the audience gets to see it. One article can be given a large stage almost immediately, while another seems to be performing in a theatre where only a handful of seats have been filled.
From my side of the curtain, I write them in much the same way. I think about the subject, write the article, edit it, publish it and wait to see what happens. Then, somewhere backstage, something I can’t see decides what happens next.
For now, the articles will continue to appear, some will find a large audience and some will apparently be performing to an empty theatre. The article is still there, of course. It’s just that somewhere between the publication button and the audience, something has managed to make most of the audience disappear.
So you may see this article and agree with it, or you might not see it at all. Now you see me...now you don’t.