The Best Marketing You Do Won't Be Read By a Human First
4 min read · Adapted from Sundae Tuesdae, Issue 8
Most of the noise around generative engine optimization treats it as a search problem. A new dialect of SEO. Schema markup, llms.txt files, content structure, citation engineering. All useful. None of it the moat.
Here's what's actually happening. When someone asks ChatGPT, Perplexity, Gemini, or Google's AI Overviews which mattress to buy, which CRM to switch to, which energy drink doesn't taste like cough syrup, the model isn't ranking ten blue links. It's making a recommendation. And like any good recommendation, it weighs the opinions of people who aren't selling you something more than the opinions of people who are.
The numbers say that out loud. Muck Rack's analysis of 25 million AI-cited links put 84% of those citations as earned media: journalism, third-party coverage, user content, with paid placements barely registering. YouTube's share of social citations across the major engines doubled in five months, to 39.2%, while Reddit's halved. The engines are listening to other people talk about you, and they're listening hardest to the people best at talking.
Every YouTube review with a clean transcript. Every TikTok demo that pulls comments. Every Reddit thread where a creator's name gets thrown around as gospel. Every podcast where a brand comes up unprompted. These are training material now. The corpus from which the machines compose their answers. Creator marketing didn't just gain a use case. It quietly acquired a second job: shaping how AI describes a brand to everyone who asks, for as long as the citations hold.
This isn't a future problem with a future invoice. It's a today problem with tomorrow's price tag already attached.
Why citation authority compounds like domain authority did
Domain authority took years to build because it rewarded consistency. A site that published useful, linkable content over time outranked a site that published once and stopped. AI citation authority is following the same shape, compressed into a shorter window because the corpus is newer and less crowded. A brand with a real, recurring footprint of earned mentions right now is building the same kind of advantage early SEO players built in the 2000s, except this time the reward is being the answer, not being link three on page one.
The brands building that footprint now will be the brands AI cites by name in two years. The brands waiting for a definitive playbook to get published will spend the rest of the decade trying to elbow into answers that already have a default. There won't be a neutral moment where everyone starts even. The models are already forming opinions.
The objection worth taking seriously
The obvious pushback: isn't this just SEO with new vocabulary? Optimize the content, get the citation, same game. It's close, but the mechanics differ in a way that matters. Traditional SEO rewards the site answering the query. Generative engines reward the source the model trusts to have answered it honestly somewhere else first, which means the brand doesn't control the page the citation comes from. A creator's YouTube review, not the brand's product page, is what gets cited. That's a real shift in where the leverage sits, and it's why a content team optimizing its own site copy for AI crawlers is solving the smaller half of the problem.
Where the actual lever is
The dot most marketing teams haven't connected yet: the cheapest, fastest, most operational lever a brand has to build that footprint is the channel it's probably already running, just measured against the wrong goal. Creator programs built for reach or engagement don't automatically produce citation-worthy content. Citation-worthy content is specific, verifiable, and exists somewhere a model can find it independent of the brand's own site: a comparison video, a long-form review, a Reddit thread where a real user vouches for the product unprompted.
Three questions worth running through a current creator program:
Does any of this content exist in a form a model could cite on its own, separate from a link back to the brand? A caption isn't citable. A detailed, specific review is.
Is the brand tracking which of its own creator content shows up when someone asks an AI engine the exact question that content was built to answer? Most brands have never checked.
Is the program optimized for engagement metrics that have nothing to do with whether the content reads as trustworthy, independent testimony to a model deciding what to recommend?
The reorientation
None of this requires a new budget line. It requires treating the creator program a brand already runs as infrastructure for a second, quieter audience: not just the people scrolling, but the models learning what to say when someone asks. The teams that notice this first won't be defending their creator budgets next year. They'll be deciding where to put the increase.
— Jeremiah Rosen, Sundae Collective