AI writes the average version of everything. Average stopped working.

Anyone can now produce a competent, on-topic, grammatically spotless blog post in about the time it takes to read this sentence. Type the topic, wait a few seconds, and out comes five hundred words that hit the keyword, cover the obvious points, and read like a real person wrote them on an average day. A lot of businesses have understandably decided this is wonderful news. The cost of "content" just fell to roughly zero.

We think it's closer to the opposite, and it's worth being precise about why. The thing these tools are genuinely brilliant at is producing the average. That's exactly the thing that stopped working.

The average is the whole problem

A language model works by predicting the most likely next word, over and over, from everything it was trained on. Point it at a topic and what comes back is, almost by definition, the statistical middle of everything already published about that topic. The median take, in the median phrasing. That's a remarkable technical achievement and a poor content strategy, because the average of what's already out there is precisely what nobody needs another copy of.

If your article is the average of the ten already ranking for a search, you've given the reader, and Google, and the AI answer engines, no reason to prefer it. It isn't wrong. It's redundant, which is arguably worse than wrong, because wrong at least gets noticed.

"Just publish more" is now the risky option

For a long time the content playbook for a lot of businesses was volume. Publish often, cover every keyword, flood the zone. Even middling pieces picked up a little traffic, and more pieces meant more of it.

That trade has inverted. Search engines have spent years getting specifically better at spotting thin, generic, produced-to-fill-a-quota content and holding it down. The AI answer boxes now sitting on top of the results happily absorb the generic answer and hand it straight to the reader, so the median article doesn't even earn the click it used to. A stack of competent, forgettable posts isn't a neutral asset gathering dust anymore. It's a signal, to the search engines and to the occasional human who lands on one, that nobody here is paying much attention.

What the machine structurally can't do

The one thing an AI can't generate is having been there. It has no clients, no failure it learned something from, no opinion it would defend in a room, no number from a real project, no story about the time the approach everyone recommended fell apart in month two. It can imitate the shape of that experience. It can't supply the substance, because the substance was never in the training data. It was in someone's actual working life.

That's the material now doing the heavy lifting. The specific figure. The contrarian view held for a reason. The honest account of what went wrong and what you changed. Not because it ticks an E-E-A-T box, though it does, but because it's the only part of a piece a competitor can't regenerate in nine seconds.

Where we use AI, and where we don't

We're not anti-AI, and it would be odd to pretend otherwise. We use it where it earns its place: pulling research together, pressure-testing an outline, getting a rough skeleton onto the page so the real work starts from something rather than nothing. What we don't do is let it supply the point of view, because the point of view is the product. The judgment about what's worth saying, and the experience that makes it worth reading, is what a client is actually paying for. The draft was never the hard part.

The bar moved. Good.

The floor for content that merely exists has dropped to nothing, which means the floor for content that does anything has risen to meet it. That sounds like bad news and mostly isn't. It rewards the businesses with something real to say and punishes the ones that were only ever filling a calendar. If you have genuine expertise, this is the best environment for content you've had in years. If you were relying on volume, it's time to change the plan.