This week I'm thinking about the thing nobody in marketing wants to say out loud. We're all racing to optimize for AI search, and at the same time we're making our content sound exactly like the AI that's supposedly the threat.

Ask yourself this. If an AI Overview can sum up your whole category in three sentences, why would anyone click through to you?

That's not a thought experiment anymore. It's just how search works now. And the response from most marketing teams has been to optimize harder. More schema, more structured data, more content built for a crawler instead of a person. Fine, as far as it goes. But it's the wrong half of the problem, and I think it's the half that's easier to solve, which is exactly why everyone's rushing at it.

Everyone's Doing the Same Technical Playbook

Structure matters, I'm not going to pretend it doesn't. Clean headers, answerable questions, markup a model can actually parse. That stuff helps you get cited. I've made these changes myself on client sites and watched the citation rate move.

But so is every single one of your competitors. Same playbook, same structured data, same moves. There are entire agencies right now selling "AI search optimization" as if it's some proprietary edge, and it's really just technical SEO with a new name and a slightly different crawler in mind. When the whole category is doing the identical thing to win the identical game, it stops being an edge. It's just the cost of showing up.

Here's the part that should worry you more than losing a ranking spot. If your only strategy is "be legible to the machine," you've built a strategy with a ceiling. You can be technically perfect and still be indistinguishable from the next technically perfect competitor. Table stakes don't win games. They just get you a seat.

The Real Opportunity Is the Thing AI Can't Copy

Here's what I keep coming back to. AI models are trained on everything that's already been published, which means the average AI-optimized article is, by definition, average. It's a blend of what already exists, smoothed out, run through a synthesis layer. So if you're writing purely to be machine-readable, you're also writing something completely replaceable — you're contributing to the exact pool of content the model is going to summarize and route around.

Meanwhile people, and honestly the models too, are getting sharper at telling the difference between content made for an algorithm and content made by someone who's actually got a take. Someone with their name on it. Someone saying the thing nobody else in the space will say. You can feel it when you read it. It doesn't hedge. It doesn't cover every angle just to be safe. It picks a side.

I've written before about how most B2B content feels generic because it's created at arm's length from the industry it's covering. AI search takes that same problem and raises the stakes. It's not just that generic content underperforms — it's that generic content is now the exact material the model was trained on. You're not just failing to stand out. You're feeding the machine that's replacing you.

That's why Google keeps leaning harder into experience as a ranking factor, not just expertise. The internet's full of stuff that sounds expert and has zero actual person behind it — content that knows the right vocabulary but has never been in the room. AI search just makes that gap more obvious, because now there's a machine sitting between you and the reader, deciding whether your content is worth surfacing at all. If a model can already hand someone the safe, consensus answer, the only reason they come to you specifically is because you've got something the safe answer doesn't have.

I'd go further. I think this is going to widen, not narrow, over the next couple of years. As more of the generic middle gets absorbed and regurgitated by AI systems, the premium on a genuine point of view goes up, not down. Scarcity works the same way it always has. The thing everyone can produce becomes worthless. The thing only you can produce becomes the whole business.

Voice First, Structure Second

I don't think this is an either-or. It's a sequencing problem. Write the thing with a real opinion and a real name on it first. Structure it for the machines after. Do it backwards — start from what'll rank and sprinkle in personality later — and it shows. People can tell. And increasingly the models can tell too, because they're getting better at weighting signals like author expertise, first-hand experience, and even how confidently a piece takes a position.

This is the same argument I made when I said your marketing team is basically a media company whether it realizes it or not. A media company doesn't win by being technically correct. It wins by having a voice people recognize without seeing the logo. AI search just makes that same logic apply to a single article instead of a whole brand.

I've seen teams try to reverse-engineer authenticity by hiring someone to "punch up" AI-generated drafts with quotes and anecdotes at the end. It doesn't work. Authenticity bolted on at the edit stage reads exactly like what it is — a costume. The voice has to be the starting point, not the finishing touch.

A few things I'd actually do differently starting this week:

None of this means ignore the technical side. Do the schema. Structure the headers. Make it easy for the crawler to find and parse what you've written. Just don't confuse that work with the actual strategy. It's the packaging, not the product.

If you're looking for guidance you can contact me here.