Here's the direct answer

Your AI-written content isn't working because AI has no information about your business — it can only recombine what everyone already published, so you get competent, correct, weightless content that says what everyone says. Content earns attention through information nobody else has: your data, your mistakes, your specific opinions, your actual client outcomes. sosSTEVIE builds content engines that start by extracting what only you know, then use AI to produce and adapt it at pace. The machine handles the volume. The information has to come from you.

Key takeaways

AI doesn't make content bad. It makes average content instant, and average content was never working.
Anything a model can write without asking you a question is something everyone already has.
Google doesn't penalise AI writing. It has always devalued content with nothing in it.
Volume without information is the fastest way to bore an audience at scale.
The bottleneck was never writing. It was having something to say, and AI didn't touch that.

The story

A man showed me his content programme with real pride, and I felt genuinely sorry for what I was about to say.

Ninety-two blog posts in five months. Before AI he’d managed about one a quarter. Now he had a calendar, a pipeline, a system. He’d cracked it.

Traffic: flat. Inquiries from content: zero. Not low — zero.

I read six posts. Every one was competent. Grammatical, well-structured, on-topic, and utterly weightless. “5 Ways To Improve Your Supply Chain.” “The Importance Of Quality Control.” Ninety-two pieces of content that could have been published by any of his competitors, because there was nothing in any of them that only he could have written.

Then we talked for twenty minutes and he told me — offhandedly, as an aside, not knowing he was saying anything — about a shipment he’d lost in 2019 and the specific mistake that caused it, and what he’d changed in his business because of it, and how he now checks one particular thing that nobody else in his industry checks.

That’s an article. That’s an article nobody else on earth can write. It was sitting in his head, unpublished, for six years, while a machine produced ninety-two pieces about the importance of quality control.

He didn’t have a content problem. He had a nobody asked him anything problem.

Everyone’s saying the same things faster

Here’s what happened to content this year.

The cost of producing a competent article went to roughly zero. Which is genuinely wonderful, and also means your competitors’ cost went to zero too. So everyone published more.

The scale of what you’re now competing with is worth sitting with. Industry estimates commonly cited by Internet Live Stats and repeated across the sector put daily output at around 7.5 million new blog posts a day, against a base of some 600 million blogs. Treat those as directional rather than precise — nobody can audit the whole web — but the direction isn’t in dispute, and it’s up.

Now the number that actually matters. Ahrefs’ large-scale index studies have repeatedly found that around 96% of pages get no organic traffic from Google at all. Not a little. None.

Read those two together. Millions of pages published daily, and almost all of them are invisible. The overwhelming default outcome for a published page is that nothing happens to it. Your ninety-two posts weren’t unlucky — they landed exactly where the distribution says they would.

Here’s the mechanism people miss. A language model can only tell you what has already been said. That’s not a flaw, that’s what it is. So if a model can write your article without asking you a single question, that article contains no information that wasn’t already public — which means, definitionally, that your competitors already have it.

You didn’t create an asset. You created a copy of something that already existed, and paid for hosting.

Meanwhile, the stuff that would actually work — your numbers, your failures, your specific contrarian view, what actually happened with that client in March — is in your head, and nobody asked.

The Google question

Let me kill this one, because it wastes enormous amounts of energy.

“Will Google penalise me for using AI?”

No. Google has been clear that it cares about content quality, not production method. It doesn’t penalise AI writing.

But that answer is comforting in a misleading way, so here’s the real version: Google has always devalued content with nothing in it. That’s not new and it’s not about AI. What AI changed is that this kind of content is now free to produce, so people produce mountains of it, and then blame the algorithm when the mountain doesn’t rank.

The penalty isn’t for using AI. The penalty is for having nothing to say — and it’s the same penalty that’s always existed. AI just removed the last thing that used to stop people: the effort of writing it.

Ninety-two posts of nothing rank exactly as well as one post of nothing. You just spent five months finding that out.

The Information First method

An extraction interview capturing what a business owner knows that nobody has ever asked them

1. Extract. Before anything gets written, get what only you know out of your head — the losses, the numbers, the opinions you’re slightly nervous about publishing, the thing you always end up explaining on sales calls. Structured interviews, recorded. Same extraction that makes you AI-readable, which is why these two belong together.

This is the hard part and the whole product. Everything downstream is logistics.

2. Find the questions. Not keywords — questions. The actual sentences your buyers say. What they ask on calls, in emails, at the moment they’re deciding. Each becomes a piece.

3. Match information to question. Where do you have something nobody else has? Those are your pieces. If you have nothing on a topic, don’t publish on it. Silence beats noise — an empty slot in the calendar is not a failure.

4. Now use AI, hard. Structure, draft, adapt, tighten, and produce the visuals. This is where it’s genuinely brilliant and saves you most of the work. It is doing the typing. It is not doing the thinking, and the moment those swap, you’re back to ninety-two posts of nothing.

5. Put you back in. The first draft will be smooth and slightly dead. Your job is the specific number, the actual name, the sentence a machine wouldn’t risk. Twenty minutes, and it’s the difference.

6. Repurpose properly. One real idea becomes a post, a video, five social pieces, a section of a sales conversation. Not by pasting it across platforms — by rewriting it for each. Lazy repurposing is visible and worse than not bothering.

7. Publish less than you can. You can now produce a hundred pieces a month. Don’t. Publish what has information in it. That’s probably four.

The seven steps of the Information First content method, ending with publish less

About brand voice

Brand voice isn’t a document of adjectives. Nobody has ever been helped by “friendly but professional, authoritative yet approachable.”

Voice is what you’re willing to say that others aren’t. It’s your actual positions. The thing you’d argue about at a conference. A voice guide that doesn’t contain at least one opinion your competitor would refuse to publish isn’t a voice guide, it’s a style sheet — and a model given a style sheet will produce something smooth, safe and forgettable, because you asked for exactly that.

I do build voice systems, and they work. They work because we start with what you believe, not how you sound.

And on AI personas: be careful. There’s a real line between using AI to scale your voice and using AI to fake a person who isn’t there. Cross it and you’ll be found out — and the finding out costs more than everything you saved. The one place you can’t afford to sound automated is anywhere someone’s trying to work out if you’re real.

Nobody asked him anything

Ninety-two posts. Zero inquiries. And a six-year-old story about a lost shipment sitting in his head that would have out-performed all ninety-two put together.

The machine didn’t fail him. It did exactly what it was asked. Nobody ever asked him.

Let’s find out what you know — $997 Diagnostic →

About Stevie

I made content for clients for years, and plenty of it was the competent, forgettable kind. I know precisely how that gets made, because I made it.

Then I lost my business and had to write something real about it, and it did more in a month than five years of tips-and-tricks posts ever did. That taught me what content is actually for.

Why Isn't My AI-Written Content Getting Any Traction?