Earned Media Is the New GEO Moat: Why PR Now Drives 25% of LLM Citations

Earned media now matters to AI visibility for the same reason backlinks once mattered to SEO: it acts as a transferable signal of trust.

That is the most important takeaway from Muck Rack’s latest Generative Pulse findings, which were recirculated widely at the end of March. The headline number got attention because it is easy to remember: journalistic and earned media sources account for roughly 25% of all large language model citations. The second number matters even more: non-paid media sources collectively account for about 94% of citations.

Those numbers do not mean PR replaces content strategy. They mean the market has been underestimating how much AI systems rely on third-party validation when deciding what to surface.

That changes the GEO conversation fast.

For the last year, most practitioners have framed generative engine optimization as a publishing problem. Write answer-first content. Add schema. tighten entity references. include facts, quotes, and FAQ blocks. Update pages frequently. All of that still matters. But Muck Rack’s data reframes the game. AI visibility is not just a formatting contest on owned media. It is also a reputation distribution contest across the wider web.

In plainer terms, brands do not only need pages that are readable by models. They need a source environment that makes the models comfortable trusting those pages.

That is where earned media becomes the new GEO moat.

Why this is bigger than a PR talking point

Whenever a PR vendor publishes research, marketers should be skeptical. The incentive to overstate the importance of media coverage is obvious.

But this finding aligns with a broader pattern that many operators have already seen firsthand. Large models routinely lean on established publications, expert commentary, reviews, roundups, and journalistic summaries when constructing answers. That is not accidental. Third-party sources offer at least three things models value:

• external validation of a brand or claim
• editorial selection that reduces source uncertainty
• language that is often more comparative and contextual than brand copy

This is why earned media performs differently from owned content inside AI systems.

Your website tells the market what you want to be known for.

Earned media tells the model what other people think you are worth mentioning.

That distinction matters because LLMs are trained and tuned to prefer sources that look independent, well-cited, and reputation-bearing. A feature in a credible publication does not just create awareness. It strengthens the probability that the brand enters the answer graph as a legitimate entity.

That is the part many GEO playbooks still undersell.

GEO has quietly become a source-graph problem

The old SEO mental model centered the page. The improved GEO model centers the source graph.

A source graph is the web of places where a model can encounter, verify, compare, and contextualize a brand or claim.

That graph includes:

• your owned site
• industry publications
• mainstream media coverage
• trade journals
• product reviews and community references
• analyst commentary
• first-party data assets cited elsewhere
• interviews, podcasts, and quoted expertise that become transcript text

In that environment, your homepage or blog post is only one node.

This is why brands that publish constantly but never earn coverage often underperform in AI answers relative to brands with fewer owned pages but stronger third-party validation. The model does not just ask, "Did this company say it?" It effectively asks, "Does the broader source environment support this company’s relevance and credibility?"

That is exactly what a PR moat looks like in an AI discovery era.

Why 25% is a massive number, even if it understates the real effect

It is tempting to read the 25% figure literally and move on. That would be a mistake.

If earned media accounts for one quarter of citations directly, its indirect effect is probably larger.

Here is why.

A media mention can influence AI visibility through several paths at once:

1. Direct citation path

The model cites the article itself.

1. Entity reinforcement path

The coverage helps the model associate your brand with a category, use case, or expertise area.

1. Query framing path

Journalistic language often supplies the comparative phrases and definitions models later reuse.

1. Link propagation path

Media coverage leads to other sources mentioning, quoting, or linking to the brand.

1. Retrieval confidence path

Even when the model cites a different source, the surrounding media footprint may increase confidence that your brand belongs in the answer.

This is what makes earned media such a strong moat. It is not just another source type. It is a trust multiplier for the whole source environment.

Muck Rack’s earlier 2025 findings already hinted at this by showing that citations can materially alter LLM outputs. The 2026 framing pushes the industry closer to the operational takeaway: if you are ignoring media presence, you are ignoring one of the strongest non-owned signals AI systems use.

Why PR teams suddenly have a budget argument they did not have before

For years, PR teams have struggled to prove attribution compared with paid media and performance channels. They could show reach, share of voice, sentiment, and branded search lift, but the direct commercial chain often looked fuzzy.

AI changes that conversation.

If earned media shapes what large language models cite, then PR affects not just awareness but answer formation. That gives communications teams a much stronger claim on digital discovery budgets.

This does not mean every press mention is gold. It means the right media coverage can now influence:

• category inclusion in AI answers
• product and vendor shortlists
• executive thought-leadership visibility
• trust in factual claims
• citation share against competitors

That is a much more measurable and strategically defensible role.

It also means marketing leaders need to stop treating PR, content, SEO, and GEO as separate silos. In practice they are becoming one system:

Function	Traditional role	AI-era role
PR	awareness and reputation	external trust signals for answer engines
Content	traffic and conversion	owned source material for retrieval and citation
SEO	ranking and click capture	eligibility, structure, and authority reinforcement
GEO	formatting for AI	multi-surface source-graph optimization

The companies that merge these functions intelligently will outperform the ones that keep optimizing each channel in isolation.

Why most GEO advice is still too narrow

A lot of GEO advice circulating right now still looks like checklist SEO.

• add FAQs
• use concise definitions
• improve schema
• publish statistics
• cite sources
• update content

Again, none of that is wrong. It is just incomplete.

The problem is that checklist GEO treats the model as if it only reads your page in isolation.

It does not.

Models and retrieval systems increasingly encounter your claims in a competitive source environment. They compare your owned page with:

• media coverage of the same topic
• competitor mentions in roundups
• third-party datasets
• commentary from analysts or practitioners
• structured snippets from community and editorial pages

In that environment, formatting excellence cannot fully compensate for source weakness.

That is the same lesson SEO learned the hard way years ago. A well-optimized page without authority signals struggled to rank. Now a well-optimized page without off-site trust signals can struggle to be cited.

So the real GEO question becomes:

How do you build a web-wide trust footprint that makes your owned content easier for models to believe?

That is where earned media becomes a moat rather than a nice-to-have.

Journalism matters because AI systems borrow confidence, not just facts

One reason journalists still matter in an AI answer economy is that they do more than report facts. They frame significance.

A news article or feature often clarifies:

• why a company matters in a category
• how it compares with peers
• what claims are credible enough to repeat
• which trend the brand belongs to
• what context a reader should use to interpret the information

Those are exactly the kinds of context layers AI systems need when composing synthetic answers.

That is why a journalistically mediated mention can influence AI output more than a self-published announcement with identical raw facts. The earned-media version arrives with editorial context already attached.

This does not mean every outlet is equal. It means the quality and relevance of the source environment matter. A citation in a trusted trade publication for your niche may be more valuable to AI visibility than a generic mention in a broad but shallow outlet.

That should change how brands think about media strategy. The goal is not only reach. The goal is source usefulness to machines that build category knowledge.

The new moat is not press releases. It is cited reputation.

Muck Rack’s broader body of research has also highlighted an uncomfortable point for many communications teams: earned media is still doing more work than press releases when it comes to generative AI visibility.

That makes sense.

Press releases are first-party framing with some distribution. They are useful, especially when they trigger pickup, create canonical facts, or help the brand define official language. But they are not the same as cited reputation.

Cited reputation happens when independent or semi-independent sources repeatedly mention the brand in relevant contexts that AI systems later recognize as trustworthy.

That implies a different operating model for PR and content teams.

Instead of asking only, "How many placements did we get?" teams should ask:

• Which placements are likely to be crawled and cited by AI systems?
• Do those placements reinforce the exact entities and claims we want associated with the brand?
• Are we earning coverage in sources models already rely on for our category?
• Are we turning original data and expertise into media hooks that travel?

This is a higher standard than vanity PR, but it is much closer to commercial impact.

What brands should build now

If earned media is becoming an AI trust moat, the tactical response is not to spray more pitches. It is to build a system that compounds third-party validation.

1. Identify your citation-critical topics

Map the prompts and answer scenarios where AI visibility matters most. Those are the themes where earned coverage will have the highest value.

2. Create data that deserves pickup

Original research, benchmark studies, industry surveys, and category analyses travel farther than generic opinion pieces. They also generate more quotable facts for both journalists and models.

3. Train spokespeople for machine-readable expertise

Strong quotes matter more now because they often become retrieval-worthy phrasing that carries into answer systems.

4. Target trusted trade and specialist outlets, not only broad reach

The most useful placement is often the one that anchors your authority in the exact topic cluster where models need confidence.

5. Reinforce earned media with owned content hubs

When coverage lands, publish complementary explainers, methodology pages, definitions, and FAQs on your own domain so the source graph points back to stronger owned assets.

6. Measure AI citation lift after media campaigns

PR reporting should now include whether media activity improved prompt-level presence, brand mentions, or source inclusion in answer engines.

This is the operational bridge between communications and revenue teams.

Why this shifts the balance of power toward brands with real expertise

There is a hopeful angle here.

For years, many digital channels rewarded sheer volume, paid amplification, or shallow optimization. AI citation systems, despite all their flaws, appear to place real weight on independent validation and source credibility. That creates an opening for brands that genuinely know something and can earn attention for it.

The shortcut mindset will still exist. Some teams will try to mass-produce fake authority through low-quality syndication or paid placements disguised as editorial presence. But the more mature answer systems get, the more likely they are to reward source ecosystems that feel coherent, repeated, and independently reinforced.

That favors brands that:

• produce original insights
• build relationships with real publications
• show up consistently in category conversations
• turn expertise into evidence, not just slogans

In other words, it favors brands willing to earn trust rather than simulate it.

The strategic takeaway

GEO is no longer just about making your own page easy for a model to read.

It is about making the broader web easy for a model to trust when it thinks about you.

That is why earned media is becoming the new moat.

Backlinks taught search engines that other sites considered you worth referencing. Earned media teaches AI systems that other people consider you worth believing.

Those are not identical signals, but they rhyme closely enough that smart operators should take the lesson seriously.

The teams that keep treating PR as a soft-awareness layer will miss one of the strongest trust signals in AI discovery.

The teams that integrate PR, content, and GEO into one source-graph strategy will own more of the answer layer.

What a modern PR plus GEO operating model looks like

The practical implication is that communications teams need a different production rhythm than they had even a year ago.

A modern operating model usually includes four connected streams.

First, a research stream.
Brands need a repeatable way to produce facts worth citing. That can be internal benchmark data, customer behavior studies, category comparisons, original surveys, or methodology explainers. The point is not to publish noise. The point is to create material journalists and AI systems both consider useful.

Second, an outreach stream.
Pitching should map directly to the authority themes the company wants to own. If the brand wants to be associated with AI visibility measurement, retail agent infrastructure, or citation analytics, the earned coverage should reinforce exactly those associations rather than scattering attention across loosely related stories.

Third, an owned-content stream.
Every major coverage win should be supported by a strong owned page that deepens the same topic. That way the media mention helps the model discover the entity, and the owned asset helps the model understand the claim in detail.

Fourth, a monitoring stream.
Teams should check whether target prompts and answer scenarios actually changed after coverage landed. If not, they need to ask whether the media source lacked authority for that use case, whether the claim was too generic, or whether the owned asset failed to convert attention into machine-legible evidence.

Put together, that creates a much tighter loop between reputation work and discoverability work.

Why this favors specialists over generic brand noise

There is another reason the earned-media shift matters. It should make the web less forgiving for vague positioning.

AI systems do not just reward the loudest brand. They often reward the brand that appears repeatedly in a clear context. Repetition inside a narrow expertise zone can outperform broad but fuzzy awareness.

That is good news for specialist companies and focused publications.

A niche B2B software vendor that consistently appears in the right trade coverage, contributes real benchmarks, and defines a category problem clearly may become more citable than a larger competitor with more ad spend but weaker narrative coherence. A specialist law firm quoted repeatedly on one regulatory issue may become more retrievable than a generalist giant. A focused research publication can become a go-to source for an AI system because its identity is easier to resolve.

This is another way of saying that PR in the AI era is becoming more like entity engineering. The job is not simply to be seen. The job is to be seen in the right context often enough that machines stop being uncertain about what you are and why you matter.

If you want to see how exposed or reinforced your brand is across AI discovery surfaces, run a benchmark at audit.searchless.ai.

FAQ

Why does earned media matter so much for LLM citations?

Because it provides external validation, editorial context, and trusted third-party language that large language models often treat as stronger evidence than self-published brand claims.

Does this mean PR is more important than owned content?

No. Owned content still provides the canonical source material. But earned media can increase trust in that material and improve the odds that AI systems include the brand in answers.

What is the main lesson from Muck Rack’s 25% figure?

That AI visibility is not only a formatting or on-site optimization problem. It is also a source-graph problem where off-site reputation meaningfully affects citation behavior.

Should brands focus on press releases or journalist coverage?

Press releases are useful, but journalist coverage is generally more powerful because it carries independent framing and stronger trust signals for models and retrieval systems.

How can a company turn this insight into action?

It should build data-led PR, target category-relevant outlets, reinforce coverage with owned explainer content, and measure AI citation lift after campaigns. A practical baseline starts at audit.searchless.ai.