Does YouTube still matter when buyers ask AI first?

Yes. For a boring reason. Buyers still have questions. An LLM still has to get the summary from somewhere.

Yes. For a boring reason.

Buyers still have questions. They still need to trust the person who will take their money. An LLM can summarise a problem. It can't sit in the work with them. And it has to get the summary from somewhere.

If you already own a clear, human answer to the thing they type, you're useful twice. Once on YouTube and Google. Once as the page a model might draw from. If you own nothing, you're on neither list.

People are treating the rise of AI answers as a reason to wait. Waiting is how you miss the only part of this still in your control. The sentence.

What actually changed

Search is less of a neat list of ten blue links than it used to be. AI Overviews. Chat windows. People asking a model before they ask a directory.

That doesn't make a weekly search video less useful. It makes the exact question more valuable.

Vague thought leadership was always weak. It's weaker now. A model can produce vague. It can't easily replace the practitioner who says, in their own mouth, why the strategy deck died in week two and what they do instead.

You with me on that?

The slop wave is real. You've seen it. Generated voice. Looped footage. Twenty minutes of nothing. That stuff fills the shelf. It also trains buyers to recognise a human who has done the work. Your job is to be obviously the second thing.

What did not change

People still type the thing that is stuck.

They still watch a person think, when the purchase is expensive and a bit frightening.

They still arrive on a call warmer if they've already heard the answer once.

Referrals still dry up. LinkedIn still buries last Thursday's post. The diary still has feast and famine in it if inbound depends on who you bumped into.

None of that was invented by ChatGPT. None of it goes away because the results page looks different.

Search estate in this climate

A search estate means you pick the handful of questions your best clients already ask, and you occupy them on purpose.

Titles in their language. Answers that sound like the call, not like a brochure. A weekly cadence so the estate grows instead of resetting.

When a buyer asks an LLM first, two things can happen. The model gives a generic answer and they still need a name. Or the model points, however loosely, at people who have written and spoken the specific version.

You can't control the model. You can control whether the specific version exists.

So the question "does search still matter" is the wrong shape. Search changed shape. Questions didn't.

What not to do with this news

Don't start a new content pillar called "AI and the future of my industry" unless a buyer is actually paying you to answer that.
Don't film a panic video about algorithms.
Don't decide the game is over and go back to cold outreach as the only plan.
Don't let an agency sell you "AI optimisation" as a separate religion. Write clear answers to real questions. That's most of the work under any name they give it.
Don't use AI to mass-produce ten videos you wouldn't say out loud. You think faster than you type. Fine. Use the tools to remove the bottleneck. Don't use them to become a slop factory with your logo on the thumbnail.

Now, can you see how this stays simple? Same method. Sharper titles. More reason to sound like yourself.

A practical week

Take one question a client asked in the last fortnight. The unpolished one.

Ask an LLM the same question. Read the answer. Notice what's generic.

Record the version only you can give. The one with the judgement. The one with the "here's where this goes wrong in week two."

Publish that. Title it in their words.

You've just done more "AI strategy" than a white paper about visibility in Claude. You've also added a plot to the estate.

Trust got more expensive, not less

When anyone can generate a tidy paragraph, tidy paragraphs get cheap.

What gets expensive is proof that a human has been in the mess. That's why a slightly imperfect video, spoken like a call, can beat a perfect article that could have come from anywhere.

I'd rather you sounded like yourself, with a title a buyer would type, than you sounded like every other "expert insights" page a model has already digested.

If that feels too small a response to a big shift, good. Small is what you can ship this week. Big is how people delay.

A last distinction

Using AI to help you write a title, clean a transcript, or find the phrase a buyer might type: useful.

Using AI to replace the judgement you were hired for: that's how you become the slop you're complaining about.

Keep the judgement. Automate the boring bits. That's how the six hours stay six.

Google is busy deciding which pages to trust. Buyers are asking models before they ask directories. That's the weather.

The work is the same work. Own the result for the thing they type. One video a week. About six hours a month. Qualified enquiries. No creator identity required.

If you want the longer walkthrough of CAT Moments and the search estate, the community is free. If you want to map the questions your calls already contain, we can do that on a call.

What question would you want an AI answer to cite you for? Start there. Not with the news. With the sentence.