What AI Search Means for Creators (And What I'm Building Now)
You've watched your blog traffic slide for months now, and every time you type a question into Google the answer is already sitting right there at the top, so there's no real reason for anyone to click through to your site anymore. If you've started to wonder whether writing content is even worth it at this point, you are definitely not the only one asking that question.
I've been sitting with that same question myself, and then I listened to an interview that gave me the clearest read on this whole moment that I've heard so far, so I want to walk you through what it actually said and then show you exactly what I'm building in my own business because of it.
The shift everyone feels but can't name
The interview was Neville Medhora talking with Tim Soulo, the CMO of Ahrefs, and you can watch the full conversation on YouTube if you want the long version. Ahrefs is an SEO data company that sits on search data almost nobody else has access to, so when Tim talks about where search is actually heading, it's the kind of read that's worth slowing down for.
His take is that this is the biggest shift he has seen in his entire career, and the way he describes it is that for years Google has been slowly keeping more and more clicks for itself by answering questions right there on the results page, and now that slow drift has turned into something sudden, because AI overviews are everywhere and a lot of people have stopped searching Google at all and just ask ChatGPT instead.
The part that actually changed how I think about this is that he doesn't believe search is dead at all, because people are never going to stop searching for things, and what has really changed is simply where they go to do that searching and how the answer gets put together once they get there.
Two kinds of content, and only one still wins
This is the idea I keep coming back to, because Tim splits all content into two different buckets and only one of them still works the way it used to.
The first bucket is all the old, crowded topics like copywriting, or email marketing, or how to start a podcast, and these are the topics that already have ten thousand articles written about them that all say roughly the same thing. AI can read every one of those articles, blend them together, and hand someone a clean answer in about two seconds without ever needing to send that person to your blog, which is exactly why competing on those topics keeps getting harder every single month.
The second bucket is the brand-new stuff, the topics that are so new that the information barely exists yet, and AI simply cannot give a good answer about something that's only a few weeks old because nobody has written that answer down anywhere, which is where the real opening is for the rest of us.
The example Tim gave is the exact thing I have been working with lately, which is why it caught my attention, because he pointed straight at MCPs. People are already searching for how to use them, and AI has almost nothing useful to say yet because the tools are still too new, so whoever writes the first genuinely helpful thing about them is the one who gets the attention and the search traffic that comes with it.
An MCP is basically a connector that lets an AI tool plug straight into another tool you already use, like your email platform or your checkout, so the AI can see your real data and actually do things with it instead of just guessing.
So the move here is to go where AI is still in the dark, which means the new tools and the new platforms and the trends that nobody has documented yet, and if you can get there early and be genuinely useful while you're there, you are the one who gets found.
The generic how-to guide is a commodity now, but the story of what you actually did with the new tool is not.
— Dr. Destini CoppWhy documenting beats generating
There was a second thread running through that whole conversation that ties right back into the first one, because both Neville and Tim landed on the same point, which is that AI can write you a generic article in seconds but it cannot go out and do something real in the world and then come back and tell you the story of what actually happened.
Neville's old example of this is a little wild, because years ago he paid people to sell bottled water on the side of the road and then wrote about the entire thing, and that story ended up getting more attention than any polished how-to guide ever did, simply because it actually happened and it was his, and people still bring it up to him to this day.
You don't need to do anything close to that risky for the principle to hold, because all it really takes is doing something real in your business and then writing down what you tried and what you learned along the way. That is the kind of content no model can ever fake, because it didn't happen to the model, it happened to you.
I have been putting that into practice in my own business too, because not long ago I built a $37 tripwire funnel in a single afternoon using AI and then wrote down the whole process from start to finish, which is exactly the kind of real build that becomes content nobody else has and that AI could never have invented on its own.
So here's what I'm actually building
This is the part I want to document for you, because I am right in the middle of it as I write this.
I've been connecting MCPs to my own business and building custom skills inside Claude to run the work I used to do by hand, and the Kit MCP is a good place to start because email is the center of everything I do. Kit is my email platform, and connecting it through an MCP means I can ask the AI to look at my real list instead of handing me textbook advice, so instead of asking what a good open rate is in general, I can ask which of my actual subject lines got the most replies and get an answer that comes from my own data rather than the internet's average.
I documented a clear example of how far this can go a couple of weeks ago, when a connector went straight into my tools and fixed a live coupon code mess for me while I was sitting out by the pool, which is the kind of real work these connectors can carry all the way across the finish line instead of just handing back to you as another to-do.
Now, I'm not going to pretend that I had any of this figured out on day one, because I really didn't, and the way it actually happened was in stages, which I think is the genuinely useful part to share with you here.
I did the whole thing by hand
I was pulling reports, copying numbers across, and eyeballing which emails worked, which is the slow, manual kind of task that ends up eating an entire Tuesday before you even notice it's gone.
I had AI do one piece of it
I handed it just one step, which was summarizing the last batch of emails, and even though that wasn't the whole job, it proved that the tool could work on my real numbers rather than a generic version of them.
I had it run the job end to end
I had it go from "look at my list" all the way through to a finished read I could act on, and the real shift at this point was mental, because I stopped thinking of it as a tool I poke at and started treating it like a process I actually own.
I taught it to sound like me
I fed it my past emails and my real edits so that it could learn my voice, and now every time I fix something I tell it exactly what I changed and why, so the next version comes back a little closer to how I actually write, and it keeps getting better the more I use it.
Most people stop somewhere around stage two and never make it to that fourth stage, which is a shame, because the reason my skills sound like me and not like generic AI is that I showed the tool dozens of examples of my own work and then kept correcting it every time it missed. Each small fix builds on the one before it, so the quality keeps climbing instead of staying flat, and that is really what it means for something to compound, where you are never starting over from scratch but always stacking on top of what you already taught it. And once one of these skills is genuinely dialed in, I can hand it straight to my team so that everyone is working from the same playbook instead of reinventing it on their own, which is something I walked through step by step in a separate post if you want to see how that side of it actually works.
With a lot of this I am still sitting at stage one myself, and Neville had a good line for that, because he said we are all roughly where the iPhone App Store was back in 2008, when the big idea was a flashlight app and nobody had imagined Uber yet. The early ideas always look small at first, but you build them anyway, because that is how you find your way to the bigger ones, and because being early is the entire advantage in the first place.
The thing AI still can't do for you
If there is one thing I want you to take away from all of this, it is that the speed of AI is completely real, but speed is no longer the hard part of the work, because judgment is the thing that actually separates good work from forgettable work now.
Tim made this point through a story about Rick Rubin, the famous music producer who doesn't play any instruments and can't work the soundboard, and yet artists still pay him anyway because he knows what good actually sounds like, which turns out to be the whole job. It works the same way with AI, because the tool will happily hand you ten versions of anything you ask for, but you are still the one who has to say which one is right and which nine are not, and your taste is the part of that nobody else can copy.
So I use AI in two different ways depending on what is at stake, because when I needed a sales page recently for something I was never going to fuss over, a first draft that was almost there was perfect and I shipped it without a second thought. But for the work that is really mine, the experiments and the builds I want my name attached to, I slow all the way down and put my own judgment on every single line, because the generic stuff can fly on its own while the work that is actually you is where you stay in the chair.
How this connects to getting found
This is where the whole thing loops back around to search, because Tim said something that should change how you think about being found in the first place. For Google, the old signal of authority was always backlinks, which are really just other websites linking back to you, but for AI answers the new signal is brand mentions, so the more that people mention you and your products in a good light across the web, the more those AI tools surface you when someone asks them a question.
That maps straight onto the first stage of the Creator Growth Flywheel, which is Attract, the stage where brand-new people first discover that you exist, and it is also exactly why the Revenue Stack Method still holds up. You start inside other people's audiences, the ones who already have the attention you want, and you get yourself mentioned there, and then you send those people to something you actually own, like your newsletter, so it really is borrowed reach first, then your owned audience, and then paid traffic last of all, once the engine is already turning on its own.
Documenting what you build feeds every part of that loop, because every real story you tell becomes one more reason for someone to mention you, search your name, and come find you again later on.
Your move this week
You don't need to go and build an MCP by Friday, and you can honestly start a whole lot smaller than that.
Just pick one new tool, use it on one real thing in your business, and then write down what happened, what actually worked, and what surprised you along the way before you go ahead and publish it. That single post is the kind of content AI cannot generate on its own, the kind that grows your name over time, and the kind that gets you cited when someone asks an AI tool who they should be learning from.
The game looks completely different than it did two years ago, but underneath all of it the rules are the same as they ever were, which is to make something people actually want, do it for real, and then tell the truth about how it went, so the only thing left to do is start now, while the new stuff is still genuinely new.
Not sure where your business is leaking growth?
Take the free Creator Growth Flywheel scorecard and find the one stage that's holding back the rest, so you know exactly what to work on first.
Take the Free Scorecard →Frequently Asked Questions
Not at all, because people are still searching just as much as before, they are simply getting their answers in new places now like AI overviews and ChatGPT, and the principle of being findable hasn't changed. What has changed is that generic, crowded topics get blended together into one AI answer, so the better play these days is being first on new topics and documenting the real things you actually do in your business.
An MCP is basically a connector that lets an AI tool plug straight into another tool you already use, like your email platform or your checkout, so it can pull your real data and take action on it. Creators should care because these tools are brand new, people are already searching for how to use them, and AI doesn't have good answers yet, which is exactly the kind of opening where you get to be the first useful voice on a topic.
Brand mentions are becoming for AI search what backlinks once were for Google, so the more that other sites and creators mention you and your products in a good light, the more AI tools surface you in their answers. That means getting yourself noticed inside other people's worlds and then sending those people to something you own still works the way it always has.
Use it where speed helps and the stakes are low, like a quick first draft you were never going to fuss over anyway, but for the content that is really you, your own experiments and your own builds, slow down and put your judgment on every line. AI can give you ten versions of anything, yet you are still the one who has to know which one is actually good, and that taste is the part nobody can copy.
It means doing something real in your business with a new tool or a new idea and then writing down what happened and what you learned from it, because AI can produce a generic how-to guide but it cannot run your actual experiment and report back. The story of what you tried is content nobody else has, and it is exactly the kind of thing AI ends up citing later on.
