The Best AEO Strategy For Creators In 2026 Is 14 Years Old

Last week I watched Jason Fitzgerald have the realization in real time during our July meeting of The Huddle. He has spent close to a decade focused on SEO for his site Strength Running: publishing optimized content, earning links by writing for other sites, choosing his own anchor text. His search traffic is now down 80 to 90 percent from its peak, gutted by AI summaries that answer the question before anyone clicks. And then he looked at the AEO playbook I handed him and said, out loud, that it was the same strategy he had 14 years ago. He is right. The skills did not change. The reader did.

Recently I ran a study with Michael Rueckert at Centium. We fed roughly 2,200 running questions into the major models and a sentiment platform to see what AI actually says when someone asks about shoes, nutrition, watches, training, all of it. We did not name brands. We wanted to know what the models believe, and which sources they lean on when they answer.

The media rankings that came back were not what I expected. Runner’s World first. Trail Runner Magazine second. Ultra Running Magazine third. iRunFar fourth. And fifth, ahead of Women’s Running, ahead of publications with staffs and budgets and decades of brand equity: Jason. One guy, one site, one podcast. His show ranked as the most-cited podcast in the entire category.

His reaction was not celebration (though that did happen later). It was a shrug and a question. What does that mean, and why should I care, if nobody clicks through to me?

That is the right question. It is also the question almost every creator is about to ask. So here is the answer, and the reason the answer is more reassuring than the panic suggests.

The way it was explained to me is that a model answers like a student taking two exams at once. There is a closed-book exam, which is everything it already absorbed about a topic over years of training. And there is an open-book exam, which is what it can pull in fresh, weighted toward what sources have said recently.

On the closed-book side, the model already “knows” the safe answers in a category because people have repeated them for years. On the open-book side, it reaches for sources it has learned to trust and checks what they say right now.

Here is why that rewards a decade of SEO instead of erasing it. The behaviors that made a source rank in Google are the same behaviors that make a source legible to a model: publishing consistently, getting cited by outlets that already carry authority, saying things that line up with what trusted sources say. Jason’s back catalog is not dead weight, but rather it is the reason the model treats him as credible. A decade of showing up, in his words “massaging” the algorithm and writing for other people’s sites, built exactly the footprint the models now reward.

It is not only what he published on his own site, either. It is the hundreds of articles he published everywhere else. Each one was, at the time, a backlink. Each one is now a place where a trusted outlet is on record associating his name with a correct answer.

I bet the hundreds of articles I’ve published elsewhere is also helping. Backlinks became sentiment?

Jason Fitzgerald, Strength Running

That question is the right one, and the answer is close to yes. A backlink used to be a vote that a ranking algorithm counted. Now the same link is a place where a source the model already trusts has said something about you, in language the model reads. The link still matters. It used to carry authority; now it carries an opinion.

So what changed? Not the work. The reader.

I asked Claude where a model’s sense of credibility comes from. The answer: it is borrowed, not earned directly. A model cannot verify whether a claim is true, so it substitutes proxies that correlate with credibility judgments humans already made somewhere else. You do not persuade the model. You change what it reads. A source reads as credible because an outlet the model already trusts has vouched for it.

That flips the whole strategy. The highest-leverage move is not optimizing your own site harder; it is transitive endorsement: getting your name and your take onto the surfaces the model already trusts. Jason spent two years as a monthly columnist at Trail Runner. He has been quoted in Runner’s World a bunch of times. That byline was worth far more than he realized at the time, because it seeded his name across every source that then quoted it. The move now is to get back into that cadence. A monthly column at a major running publication does more for his standing in the models than another year of posts on his own domain, because it is the trusted outlet vouching for him, not him vouching for himself.

The economics have flipped, too. He used to get paid a few hundred dollars an article to write those columns. Given what the placement is now worth to his own business, where the highest-leverage product is a training plan that sells with almost no marginal cost, the column is worth writing even if the check gets smaller. Possibly worth writing if there is no check at all, though don’t tell that to the publications!

And the surface area compounds. When Michael analyzed my own footprint, the thing he flagged was not any single piece of content. It was that I touch a lot of things in a lot of places. I write here, on my website, publish on YouTube, Twitter, Threads, Instagram, and LinkedIn. The model needs many places to find you before it trusts you. YouTube, transcripts on, punches above its weight. Being a guest on other people’s shows still works, because the links and the mentions stack. The operator who has built trust across a dozen surfaces for years is holding the exact asset that matters, and most of them do not know it yet.

Here is where it gets interesting, and where I think a real market is about to appear.

If you are publishing on a trusted outlet on a regular cadence, you have editorial room to weave in the brands you actually use. Not a coupon code or a link to click. A single sentence in a roundup that mentions a product by name, in context, the way that happens in real editorial all the time.

We do not yet know what that sentence is worth. But run the math. If one editorial mention nudges a model toward recommending a brand, and that brand is doing nine figures a year, a single sentence could possibly be worth five figures in downstream sales to them. Nobody prices this yet. In a year, one of two things will be true: we will understand exactly what an editorialized mention is worth, or it will be so commoditized it stops mattering. Either way, the window to be early is now, and it is measured in months, not years, because the models’ recency bias favors fresh content and the field is still wide open.

It is worth saying the uncomfortable part more directly. Some of this reads as gaming a system. Content has been a game for at least five years. The difference is that now you know the rules, so you might as well play instead of watching other people play. (I said almost exactly this to Jason in our meeting).

• The skills transfer. A decade of SEO fundamentals (consistent publishing, earned links, saying what trusted sources say) is what makes a model treat you as credible. The panic is mostly people realizing they never built the footprint, not that the footprint stopped working.

• Credibility is borrowed. You do not persuade the model, you change what it reads. Get onto surfaces it already trusts through bylines, guest spots, and citations, rather than only optimizing your own site.

• Surface area is the asset. The model needs many places to find you before it trusts you. Expand the footprint: put video on YouTube with transcripts on, keep guesting, keep publishing in more places.

• Ranked lists are the unit of recommendation. “Best of” and “top” formats get pulled disproportionately. Build them, put the year in the title, and keep them fresh.

• The one-sentence mention is unpriced. A single editorial mention on a trusted outlet may be worth far more than it looks. Get in before the market figures out how to price it.

Jon Levitt is the host of For The Long Run, founder of the Long Run Labs Network (35+ shows, ~1M monthly downloads), and co-founder of The Huddle. This newsletter covers the business of creator partnerships, sponsorship strategy, and what the data actually shows, in addition to a weekly article from that week’s Long Run Labs Podcast.