As we learn more and more about what influences LLMs and what gets citations/brand mentions, AEO has become not just an SEO channel, but one we recognize is touched by multiple disciplines - content, PR, social, etc.
The question now is, does paid editorial get cited by AI?
Like any good marketer, I started my investigation into this topic some time ago with a simple Google search. Here’s what the AI Overview told me:
This response immediately set off alarm bells in my head….because this directly contradicted my experience.
You see, many, many months ago, when the AEO/GEO sphere was still stuck on things like schema, we started to hatch a plan. We’d noticed that news sources and third party blogs were HUGE for citations in AI and they influenced responses even more than owned media. We realized as a team, a more unified approach was going to be needed to help one of our clients succeed.
We started meeting with different departments, walking through what the data showed us and pitching a cross-functional strategy. When we brought it to PR, complete with examples of journalistic or new sources that were being cited for this client, the response was a little more cool than just lukewarm, They said essentially, well yeah, we don’t disagree with you, but just so you know, like half of those examples are actually paid editorial and that’s a completely different department.
Yikes.
But after learning more from this PR department, something became abundantly clear to me, which hopefully is a no brainer for everyone else. Different marketing disciplines are just as nuanced as my own, and it pays to take the time to get to know the people who are in the weeds everyday. Because they probably have the answers you’re looking for.
The instinct that got us this far is the same one that's tripping us up
SEO people are good at a specific kind of thing: treating huge amounts of web content as uniform, measurable data. Crawl it, count it, classify it by domain and pattern, rank it. That instinct is correct — it's why SEO works. Google's whole system was built to be crawled and measured that way, so the tool fit the problem.
AI citation behavior is not that problem, or at least not only that problem. LLMs, very simply, are meant to mimic the human thought process using math. This means that citations and sources are not always cut and dry, perfectly logical. When it comes to journalistic type sources, it can pull from PR, from editorial standards, from advertorial disclosure rules, from the specific and sometimes genuinely bizarre ways individual publishers decide to sell native advertising. That's not data you can see from the outside by crawling a domain. It's institutional knowledge — the kind that lives in the head of someone who's spent years pitching editors, negotiating around embargoes, and learning which outlets blur the line between "sponsored" and "editorial" and which ones don't.
So when SEO instincts get pointed at that world without anyone who actually knows that world in the room, you get mistakes. Not because anyone's bad at their job but because the tools and thought process that make you good at one thing is exactly the blind spot on the thing next door, and it doesn't look like a blind spot from where you're standing. It looks like the same job.
A small, telling example
So back to that AIO.
“Paid editorial gets cited less than 0.3%... and earned media gets cited upwards of 84%.” The more I looked into it, the more I saw this stat being repeated over and over. It's become one of those numbers people cite without checking, the way "you only have 8 seconds to make an impression" got repeated for a decade before anyone asked where it came from.
My gut was telling me this wasn’t right. I knew paid editorial got cited. Now i just needed to prove it.
SO I dug into the study it was based on. The methodology behind that number, from what I could tell, classified content mostly by domain. ie: is this a news site, is this a corporate blog, is this a press release. Reasonable, at scale.
Except: a lot of major outlets run sponsored content directly alongside their regular editorial coverage, on the same domain, sometimes with URLs that look nearly identical unless you know the specific pattern to look for. A domain-level classifier will wave all of that through as "journalism," because at the domain level, that's exactly what it is. This was EXACTLY the trap I had fallen into ages ago when talking with that PR team the first time. My SEO logic failed me, and I suspected that it had here as well.
So I began running a few tests. I had ChatGPT generate me a handful of prompts and ran them through a few different LLMs by hand, just to see what I got.
Just after running a few prompts I got what I was looking for.
Forbes.
The study that the 0.3% stat came from classified Forbes as a journalistic source. But when I actually opened the link:
Bam. Paid editorial.
This is where general domain sorting might fail you as an SEO. Because looking at the url for this article https://www.forbes.com/sites/deloitte/2026/07/22/with-the-rise-of-agentic-has-saas-seen-its-moment/ vs a regular non paid article: https://www.forbes.com/sites/sofiachierchio/2026/09/04/ai-isnt-replacing-stock-brokers-its-making-them-better/ there really isn't a way to tell without opening the links and looking for a paid disclosure.
So I ran more tests. I analyzed over 16,000 citations and guess what?
I got 0.32% from paid sources.
Whoops. So maybe I’m wrong?
I kept digging.
I had an agent go in and check each of the links that pointed to mixed-media model type businesses like Forbes, who run paid pieces right alongside non paid pieces.
The percentage jumps from 0.32% to 0.76%.
Ok we’re getting somewhere.
So then I did even more digging. What about Forbes Advisor? Other sources like it? Affiliate linking that isn’t supposed to affect recommendations, but do we really trust that completely? No single company bought that placement, but the page only makes money if you click through and buy and that's a real financial incentive shaping what gets recommended, even without a sponsor's name on it
We get 1.30%.
Not a crazy big number. But it's 4 times more than what that initial stat reported.
And it got me thinking…if the assumptions were made about sources based on domains in this study, what other assumptions were made?
Overgeneralization is a great way to get other departments to dislike you
Another stat often cited from the study is that 84% of AI citations come from “earned media”.
But calling anything not owned media “earned media” is not logic that holds water. Some sources, like Wikipedia, while not not earned, are definitely not in the same category as other traditional earned media. No PR team is going to take you seriously calling it that.
Instead, in this age of AI search, these are more like Authority Media or Knowledge Graph Sources. Because they aren't marketing channels—they are arbiters of truth. You don’t earn your way to a Wikipedia page from pitches and press releases, but through true merit.
With this in mind,, I categorized the citations to more accurately reflect their true nature, and here’s what I found:
So,after separating out true paid content, owned domains, government sources, academic papers, hospitals, nonprofits, consultancies, and small vendor blogs….71% of everything cited still can't be responsibly classified without opening each one individually.
Recognized journalism, once you actually separate it out honestly, was 3.82%.
Everything else in that the "earned" bucket was other things wearing journalism's clothes: government pages, academic papers, hospital patient-education content, consultancy thought leadership, nonprofit research, and …. a huge mass of small SaaS vendors' own blog posts ("Remote Work Productivity: 15 Statistics You Need to Know," published by an employee-monitoring software company, etc) that look exactly like independent research but are actually content marketing for a product.
Basically, the harder you look, the more any single-number breakdown falls apart if you’re trying to classify citations based on categories quickly and programmatically.
Now of course there are caveats. My sample size is much smaller than the initial study. I'm just one person running these tests on my couch in my pjs once my kids are in bed.
But I think my point still stands. We don’t know what we don’t know. If we truly want to understand the nuance of what gets cited in AI, we have to break down the silos of marketing and talk to other disciplines. I would’ve never known to dig deeper if I hadn’t had the experiences I did with the PR teams.
What breaking down the silo actually looks like
"We need cross-functional collaboration" is the kind of thing that sounds true and changes nothing, because nobody knows what to do with it on a random Tuesday when you have a completely full to-do list. So here's the version I'd actually act on:
Before you classify or count anything about a domain's content type, ask the people who deal with that outlet for a living. A PR person who pitches Forbes regularly knows in about five seconds whether a given URL pattern is a staff byline, a contributor, a paid program, or something else entirely — because it's their job to know which door they're knocking on. That knowledge doesn't show up in a crawl. It shows up in a Slack message to the right person.
That's a small ask. It's also, I think, the actual shape of what "AEO" needs to become if it's going to be a real discipline and not just SEO wearing a new acronym: less "how do I apply my existing playbook to this new surface," more "whose expertise am I missing before I say something confident."
So the answer to the question that started this all:
“Does paid editorial get cited by AI?”
Well like everything in marketing…it depends.
