A few years ago, producing something polished was proof that you had done the work. Clean copy, a decent thumbnail, a tidy newsletter. It all signalled effort, and effort signalled that a real person cared. That link has quietly broken. Today anyone can generate a passable article, a slick image or a confident sounding post in under a minute, and most of it looks exactly like the work that used to take a week.
So the old signals of quality no longer tell us who is real. When everything is smooth, smooth stops meaning anything. What people are actually starved of now is the opposite. They want to feel that a specific human, with a name and an opinion and something at stake, stood behind the thing they are reading. That feeling is what I have started calling human signal, and I think it is the scarce asset of this decade.
Human signal is the part of your work that only you could have made. In a feed full of machine output, it is what makes a reader think a real person decided this. You build it with specific opinions, lived examples and a consistent voice, not with more polish.
Polish used to be the proof. Now it is the noise.
For most of my working life, the way to stand out was to be more finished than the next person. Better structure, fewer typos, a sharper design. Buyers used those cues to guess at competence, because producing them took time and skill. The trade held for decades.
Generative tools have collapsed the cost of that finish to almost nothing. I can ask a model for a tidy essay on any topic and get something that reads well and says very little. The surface is perfect and the centre is hollow. When the whole feed looks like that, polish flips from a signal of care to a warning sign. If it is a little too smooth and a little too general, I now assume no one really thought about it.
What actually reads as human
The things that survive this shift are the things a model cannot invent on your behalf. A real opinion that could annoy someone. A specific example from a project that genuinely happened, with the awkward detail left in. A way of phrasing that you would recognise as mine even without my name attached. These are not flourishes. They are evidence that a person made a choice.
That awkward detail is usually the part a model would smooth away, and it is usually the part worth keeping. Being awkward is one of the superpowers I have identified in myself, and it is something I will be writing on a fair bit here on HeyJon. It is only awkward if you make it awkward and sometimes it has to be awkward. The specific, slightly uncomfortable truth is exactly the thing no machine will volunteer for you, and it is often the strongest signal in the whole piece.
I test my own work with a simple question. Could a stranger, or a model with no access to my life, have produced this exact piece? If the answer is yes, the signal is weak and I have really just decorated an average. If the answer is no, because it carries a view I would defend and a story only I could tell, then it is worth publishing. This is the same instinct I bring to how I think about podcasting, where the whole value is a real person talking in a way that cannot be faked into existence.
When anyone can generate anything, the one thing you cannot download is the fact that you meant it.
Use the machines, keep the authorship
None of this means going back to doing everything by hand. I use these tools all day. They draft, they gather, they take a boring first pass off my plate so I can spend my attention where it counts. The mistake is not using them. The mistake is letting them be the last hand on the work.
So I treat the model as a fast, tireless assistant and myself as the editor who has to answer for the result. It can give me ten openings. I choose the one that sounds like me and rewrite it until it does. It can summarise a topic. I add the example, the caveat and the opinion that make it mine. The output speeds up, but the judgement stays human, which is exactly the balance I try to protect in my own process. If you want a longer look at how I keep producing without burning out, I wrote about that in Record Once. Publish for a Month.
Advisory and delivery, one system
Human signal is easy to say and hard to keep up alone, because the honest, specific work is also the work that takes real time. That is where the split I rely on helps. I do the thinking, the point of view and the voice. My team at TIM Africa handles the delivery around it, so the human parts do not get quietly automated away just to hit a schedule. The tools make us faster. They do not get to decide what we actually believe, and they never sign the work.
Questions people ask
Is human signal just personal branding by another name?
No. Personal branding is how you package yourself. Human signal is the residue of real work, real opinions and real people that others can feel in what you make. You cannot fake it into existence with a template, and you cannot generate it on demand.
Should I stop using AI tools to protect my voice?
No. I use these tools every day. The point is not abstinence, it is authorship. Let the machine draft, gather and speed you up, then put your judgement, your examples and your name on top so the result still sounds like a person decided something.
How do I know if my content still has a human signal?
Read it aloud. If a stranger could have written it, or a model could have produced it without you, the signal is weak. Strong signal carries a specific opinion, a lived example and a way of saying things that only you would choose.
Where does a podcast fit into this?
A conversation is hard to fake and easy to recognise. Your voice, your pauses and how you handle a real guest all carry signal that text struggles to match. That is why I keep recording, and why my team helps turn one honest conversation into a month of human output.
Want your work to sound like you again?
Strategy from me, delivery from my team at TIM Africa.