Your PR Budget Is Now an AI Visibility Budget. Nobody Told You That Yet.

A client came to me a few months ago frustrated that ChatGPT wouldn't mention them, even for a category they genuinely lead in. Their website was clean. Their blog was active. Their case studies were some of the best I've seen in their industry. On paper, they had every reason to show up in an AI answer. When we actually pulled their citation data, they were almost invisible. Meanwhile a smaller, less polished competitor was getting named constantly.

The difference wasn't the website. It was that the competitor had spent the last year getting written about, quoted in trade publications, mentioned in a couple of "best of" roundups, and covered by a journalist who follows their category closely. My client had spent that same year writing about themselves, on their own domain, and nowhere else. That's the whole gap. I've since seen the same pattern often enough that I no longer think of it as a fluke. It's just how these systems work, and most marketing budgets haven't caught up to it yet.

The Number That Finally Made This Click for Me

I'd suspected this for a while, but a research release from 5W, the AI communications firm, put a figure on it that made the pattern impossible to ignore. Their study, built on Muck Rack's analysis of more than one million AI prompts, found that 85.5% of AI citations reference earned media sources rather than a brand's own website. Separate research out of the University of Toronto found that AI engines cite earned media roughly five times more often than brand-owned content. Two different research teams, two different methods, landing on almost the same conclusion.

Once I saw that number, my client's situation stopped looking like bad luck and started looking like a predictable outcome of where their budget had been going. They'd invested heavily in owned content and almost nothing in getting covered anywhere else. In an AI-first discovery environment, that's a strategy built entirely around the 15% of citations that matter least.

Why AI Models Lean This Hard on Earned Coverage

Earned coverage works like a chain of trust: a journalist vouches for a story, the story vouches for a brand, and only then does the model treat that brand as credible.

The logic actually makes sense once you think about it from the model's side of the table. A large language model is trying to answer a question credibly, and it has learned, the same way a person would, that a brand's own description of itself is the least reliable source available. A press mention, an analyst note, a third-party review, or a journalist's writeup has already passed through someone else's judgment before it ever reached the model's training data. Your own homepage hasn't passed through anyone's judgment but your own marketing team's.

Muck Rack's ongoing research backs this up in a way I found genuinely useful. Their most recent study, covering more than 25 million links across ChatGPT, Claude, and Gemini, found that earned media accounts for 84% of all AI citations, while paid and advertorial content barely registers at 0.3%. Journalism specifically made up 27% of everything cited. As Muck Rack co-founder Greg Galant put it, describing the consistency of the finding across three separate editions of the study, "earned media is what AI trusts."

That's a fairly blunt way of putting it, but I haven't found a better one. We touched on a related piece of this when we wrote about why AI keeps citing the same 30 websites over and over. It isn't that those 30 sites are the best writers in their field. It's that they've been referenced, quoted, and linked by enough other credible sources that the model has learned to treat them as trustworthy shorthand for an entire category.

The Stats That Changed How I Brief Clients

I pulled together the numbers that mattered most to me once I started digging into this, because I wanted something concrete to bring into planning conversations instead of a vague feeling that "PR matters more now." Here's what the research actually shows.

The math isn't close. Earned coverage outweighs owned content by roughly six to one when it comes to what AI models actually cite.

The conversion stat is the one I bring up most in client meetings, because it reframes the entire conversation. A visitor arriving through an AI-cited mention isn't just a nice-to-have brand impression. They're converting at close to five times the rate of a typical organic visitor, largely because by the time they land on your site, an AI model has already vouched for you.

You can't buy your way into an AI answer. You have to earn your way in, and someone else has to do the vouching.

The Gap Between What PR Teams Pitch and What AI Actually Reads

Only about 2% overlap separates the journalists most PR teams are pitching from the journalists AI models are actually citing.

Here's the part of the research that genuinely surprised me. Muck Rack found that the journalists PR teams pitch most heavily overlap with the journalists AI models actually cite by only about 2%. Almost none of it lines up. Most communications teams are still optimizing their outreach for the outlets that move the needle on traditional reach and SEO backlinks, while AI models are quietly pulling from a different, often more niche and specific set of sources entirely.

That's not a reason to abandon traditional PR targeting. It's a reason to run both lists side by side. We laid out the broader case for this shift in why digital PR is the most underrated SEO strategy in 2026, and this data is the clearest evidence yet that the argument holds for AI visibility just as much as it does for rankings.

What I've Started Telling Clients to Actually Do About It

• Audit where your current citations come from, not just how many you have. A brand can look fine on paper and still be almost entirely absent from AI answers if every citation traces back to its own domain.
• Spread the same idea across more than one outlet. The 325% lift from multi-outlet distribution isn't about writing more content. It's about placing the same strong idea in front of more third-party editorial judgment.
• Build relationships with the specific journalists and outlets your category's AI answers actually cite, not just the ones with the biggest traditional audience. Those two lists overlap far less than most teams assume.
• Keep press releases lean, factual, and bullet-heavy. Muck Rack's research found cited releases carry noticeably more objective statistics and structured formatting than releases that get ignored.
• Treat recency as a real lever. A meaningful share of AI citations come from content published within the last several months, so a single old press hit won't carry a brand's visibility indefinitely.

Why This Is a Budget Conversation, Not a Tactics Conversation

The reason I keep coming back to this framing is that I don't think most marketing leaders have connected the dots yet between their PR line item and their AI visibility. For years, earned media got budgeted as a brand and reputation expense, nice to have, hard to attribute, first thing cut when a quarter gets tight. That math no longer holds. If 85% of what AI models cite about your category comes from earned coverage, and AI-referred visitors convert at multiples of your organic rate, then the PR budget isn't adjacent to performance marketing anymore. It's feeding the exact channel that's converting best.

ChatGPT alone crossed 800 million weekly users by October of last year, roughly double what it was eight months earlier. That growth curve isn't slowing down, and every one of those sessions is an opportunity for your category's earned coverage to either put your name in front of someone or leave it out entirely. The brands treating this as a real budget line, not an afterthought, are going to compound that advantage for years while everyone else is still asking why their beautifully written blog isn't showing up anywhere.

The Mistake I See Most Often, and It Isn't a Lack of Effort

What gets me is that most of the marketing teams I talk to aren't lazy about this. They're often working hard, publishing consistently, and genuinely proud of the content they're putting out. The mistake isn't effort. It's where that effort is aimed. Owned content is the part of the funnel you can fully control, so naturally it's where most budgets and most hours go. Earned coverage feels slower, less predictable, and harder to forecast in a quarterly plan, so it quietly gets deprioritized even by teams who know, intellectually, that it matters.

I used to run my own content the same way. I'd rather ship five polished posts on our own site than spend the same time and energy trying to get one paragraph placed in someone else's publication, because the five posts felt like guaranteed output and the one placement felt like a gamble. The data has forced me to rethink that trade entirely. Five posts on our own domain, by the numbers above, are competing for roughly 15% of what AI models actually cite. One solid placement in a trusted outlet is competing for the other 85%. Once you see the math that way, the "safe" choice stops looking safe.

None of this means owned content stops mattering. It's still where you house your deepest thinking, your original research, and the frameworks that make earned coverage possible in the first place. A journalist can't quote a point of view you never published anywhere. But owned content in this environment is best understood as the raw material for earned coverage, not the finish line. The finish line moved.

I went back to that client and we rebuilt their plan around getting covered, not just publishing. Six months in, their citation volume has moved in a way their content calendar alone never touched. That's really the whole lesson here. The best piece you'll ever write about your own company still won't do what one good piece written about you, by somebody else, can do in this environment.

If you want a deeper walkthrough of how to actually earn those citations once your outreach is pointed in the right direction, we put the practical steps together in how to get AI Overviews to cite your content instead of your competitors.