Fire the AI ghostwriter. The engagement numbers just handed you the receipt.
If you have spent 2026 pasting half-thoughts into a chatbot and posting whatever comes back, I have bad news and good news. The bad news is the platforms can smell it. The good news is there is a far better job for the same tool.
Angus Sewell (Angus the Nontechnical) describes a six-step stack that studies your best posts and your rivals’, turns the pile into a findings brief, and tells you what to tweak every few days. His rule is blunt: “You still write every post.” I think that is exactly right, and the data now proves it in a way it did not a year ago.
Here are the numbers. Originality.AI found more than half of long-form LinkedIn posts were likely AI-written by 2025. Predictably, the reward for joining that flood collapsed. Purely AI-generated posts pull around 45% fewer interactions than human ones, and in March 2026 LinkedIn’s own authenticity update began actively demoting generic, template-shaped content, its ranking model built to detect and bury the bland stuff. Views are reportedly down about 50% year on year, and the accounts surviving that cull are the ones that still sound like a person.
So the ghostwriter play is a race to the bottom you cannot win. The coach play, using AI to analyse what works and then writing it yourself, sidesteps the whole problem. You keep the voice the algorithm is now rewarding, and you offload the boring diagnostic bit.
One point of pushback, though, because Angus’s stack studies your rivals too. Feed a model everyone’s winners and it will nudge you toward the average of what already works, which is how a whole feed ends up sounding like one nervous influencer wearing forty faces. A findings brief should tell you what your audience responds to, not turn you into a tribute act for the loudest account in your niche. Use the coach to sharpen your angle, not to sand it down to match the pack.
Treat AI as the analyst who reads the game tape. You are still the one who has to play.
Worth knowing: LinkedIn’s 2026 crackdown cut the reach of engagement-bait polls so hard they now average a 0.07% engagement rate, which is a rounding error dressed up as a feature.