AI Content Isn't Dying. Average Content Is.

Facebook told me I had eight posts.

Not eight posts that week. Eight that counted, out of everything I'd published. The AI-assisted ones weren't in the tally.

That was the moment an abstract trend became a concrete number on my own account. And it made me go and actually check what's happening, rather than repeat the take everyone else is repeating.

The take everyone's repeating is "AI content is dead, human content is back." That's wrong. What's actually happening is more useful to know.

The Flood Is Real

Start with scale, because the scale is genuinely hard to sit with.

AI-generated articles overtook human-written ones in November 2024. By then 50.3% of new articles were AI-created. Roughly 57% of all online text has now been generated or translated using AI. And 74.2% of newly created web pages contain some AI-generated content — that's from an Ahrefs study of 900,000 pages.

The production side moved just as fast. Non-AI blog creation fell from 65% to 5% in two years. 97% of content marketers plan to use AI for content creation this year, up from 90% last year. Nearly every blog post published now involves a machine somewhere in the process.

So the flood is not a prediction. It already happened.

I wrote about a version of this in Building Isn't the Moat Anymore. Distribution Is. — when the cost of building collapses, leverage moves to trust and attention. Same dynamic here. When the cost of publishing collapses, leverage moves to credibility and specificity.

The Part Everyone Gets Wrong

Here's where the popular take falls apart.

Google does not penalize AI content. Ahrefs looked at 600,000 pages and found a near-zero correlation — 0.011 — between AI content and ranking penalties. Effectively noise.

And if you look at what's actually ranking, the picture is the opposite of the hot take. Of content in top Google results in 2026:

• 0.4% is entirely AI with minimal or no human editing
• 58% is partially AI with real human editing
• 41.6% is fully human-written

Read that middle number again. The single most common profile of a page that ranks is AI-assisted and human-edited. Not human-typed. Fully human content still holds a solid 42%, but it isn't winning by default and it isn't staging a comeback.

So the machine didn't get banned. What got exposed is the stuff with no human judgment applied to it — and that category is now vanishingly rare at the top, under half a percent.

The signal isn't who typed the draft. It's whether anyone thought about it.

What the Platforms Are Actually Doing

The platform changes are real, they're just not AI detection.

In March 2026, LinkedIn's engineering lead announced they were replacing a patchwork of specialised ranking models with a single LLM-powered retrieval and ranking system. Search Engine Land confirmed the shift. The new system reads a post alongside the author's professional profile and the reader's interaction history, and predicts whether it's relevant to that specific person.

The old system scored a list of separate engagement signals, which meant you could game it by stacking those signals. The new one evaluates meaning. If your reach fell off a cliff around March or April, that's the reason — the 2024 playbook of engagement bait and hashtag strategy stopped working, and no amount of running it harder brings it back.

Google's behavioural signals work the same way. Dwell time, scroll depth, return visits. A post that looks credible but says nothing new trains the algorithm fast: people land, skim, bounce, never return. That compounds and drags the domain with it.

Then there's the audience, which is the part most people ignore. 82.1% of US consumers say they notice AI content. 52% reduce engagement with content they believe is AI-generated. 62% are less likely to engage with it on social specifically. The algorithms didn't invent this preference. They learned it by watching people, and now they enforce it.

You can't fix that by bolting a personal anecdote onto a templated post. What you can't fake is whether someone actually stopped and cared.

So What Actually Wins

Anything that can't be produced at scale without you specifically.

A point of view. A story from a particular call with a particular client. A prediction you'd be willing to be publicly wrong about. A framework that came out of a failure rather than out of reading about failures. What you tried, what broke, what you changed.

Google formalised this with E-E-A-T, and Experience is the first word — first-hand, lived, in-the-room experience. A model can't generate that. You have to have been there.

This maps to Skills Are Not Important Anymore. Here's What Is. — AI commoditises the execution layer. What's left is judgment, lived experience, and a perspective specific enough to be wrong.

None of it is complicated to produce. It's just slower than prompting, which is exactly why it stays scarce.

Where I'm Taking This

The obvious move, if experience is the moat, is the format that's hardest to fake.

Raw video. Unpolished, in the room, me talking about what I'm actually doing that week. No script laundered through a model, no stock footage, no AI voice. It's the format where "were you actually there" is answered in the first two seconds — and it's the one I'm committing to.

As I explored in No One Is Safe From AI. Here's What Survives. — taste and judgment compound. The more average content floods the feed, the more a real perspective cuts through.

The Play

Publish less. Mean more.

Use AI. Genuinely — 58% of top-ranking pages do, and refusing to on principle is just choosing to be slower. But use it on work that started with something only you know, and edit it like your name is on it. Because it is.

The thing that died isn't AI content. It's content with nobody home.

Pick one thing you know from direct experience this week. Write about it. Don't summarise, don't balance perspectives, don't hedge for search intent. Say what you actually think, back it with what you actually saw, and put your name on it.

That's the moat. It's surprisingly defensible — because most people won't do it.

---

If you're figuring out how AI fits into your content or business strategy, book a 30-min AI discovery call and I'll walk you through what I'd do in your position.