Earned media for AI: how PR and citation strategy work together

91% of AI brand mentions come from earned media, not the brand's own website (Idea Grove). That single statistic should reorganize how a marketing org thinks about PR. Digital PR is no longer a link-building line item that sometimes lifts authority. It is the dominant input to whether ChatGPT, Perplexity, and Google AI Overviews mention your brand at all. The work is the same writing, pitching, and follow-up most PR teams already do, but the targets, the placements, and the measurement change.

Why 91% of AI brand mentions come from earned media

The Idea Grove finding is consistent with how retrieval-augmented generation actually works. ChatGPT, Perplexity, and Google AI Overviews assemble answers from passages retrieved across the web, and the model needs to see your brand named in a source it trusts before it can recommend you. A brand's own site is one possible source, but the reranker treats first-party content as inherently promotional and weights it accordingly. Third-party coverage carries more retrieval weight because it doubles as an authority signal and an entity-disambiguation signal, naming your brand alongside the topic entity and category. Parse tracks AI visibility across ChatGPT, Google AI Overviews, and Perplexity, and across the dataset the pattern repeats: the brands that appear in AI answers are the ones whose names show up across many third-party domains the model trusts, not the ones with the deepest owned-content libraries.

The 0.664 correlation that should reshape PR strategy

The most useful number for the PR conversation comes from ConvertMate and Digital Bloom's 2025 analysis: brand web mentions have a 0.664 correlation with AI visibility, while backlinks correlate at 0.218 (Digital Bloom; Wellows). That is roughly three times the predictive power. Ahrefs' independent analysis of AI Overviews visibility reached the same conclusion through a different method: branded mentions, not link equity, are the strongest external signal for AI inclusion (Ahrefs). The operational shift is concrete. A linked feature in a tier-three blog used to be the standard PR placement metric. For AI visibility, a clean named mention in a tier-one publication with no link can outperform a linked feature on a low-trust domain, because the AI signal is the named brand entity in an authoritative context, not the followed hyperlink.

Which earned media sources actually drive AI citations

The citation graph is not evenly distributed, and the surface area worth pitching is narrower than most PR plans assume. Across platforms, Reddit, Wikipedia, YouTube, and a small set of news and review domains dominate (Profound; Search Engine Land). Parse's own analysis of the source domains AI cites most finds YouTube edging out Reddit at the top of the list. The setup matters: the dominant source flips by platform, so a single-platform pitch list will under-perform.

Source: Profound, Otterly.

The right pitch list combines tier-one news and category-defining verticals (citable across platforms), trade publications and review sites (citable in B2B contexts), and community surfaces like Reddit and Quora that PR has historically ignored (CMSWire). The exact mix depends on your category. See which domains AI models cite most for the source distribution by topic. No plausible mix is dominated by your owned blog.

If you want to see which sources shape AI answers about your brand, run a free brand check — it takes a minute.

What AI citation rewards that traditional PR coverage does not

Three things change about what makes a placement valuable. First, named brand mentions in body copy beat headline-only mentions. The retrieval chunk is usually a 200–800 token passage, so a paragraph that says "Brand X, the Y for Z, has done ABC" earns the citation even with no link, while a brand-in-headline placement with a body that pronouns through the rest of the piece often loses to a less prestigious outlet with denser entity coverage. Second, structured data beats prose. PR placements in posts that include a comparison table, a stat box, or a numbered list earn citations at roughly 2.5× the rate of prose-only features (Digital Bloom). Third, time of publication matters more than it did. ChatGPT's reranker favors pages updated in the last three months, and Perplexity's crawl prefers content less than a week old. A two-year-old feature in a top publication still carries entity weight, but a recent feature in a mid-tier outlet often wins the slot.

How to pitch high-citation domains

Most PR pitches still optimize for journalist attention rather than for AI citation patterns. The shift is incremental but specific. Lead with a fan-out-friendly hook: an angle that maps to a question an AI model might fan out into, like "the three sources buyers actually trust when comparing helpdesk software" rather than a generic "we launched a thing" story. Pair the pitch with a one-page data sheet of named stats, sample sizes, and methodology, the same artifacts that drive citation in long-form content drive it in news articles too. Pre-write the brand-entity sentence you want to appear in copy: "Brand X, the AI visibility platform tracking ChatGPT and Perplexity, found Y." Most working journalists will paraphrase but keep the structure, and that structure is what the reranker chunks. For a closer look at the unit-level pattern, see the answer capsule technique. And keep one short note in the follow-up: ask for a body-text mention with a brand alongside the category entity. That is the asset.

Vertical patterns: where earned media earns its slot

The right earned-media targets shift by vertical, and the differences matter for budget allocation. In B2B SaaS, the citation graph leans on G2, Capterra, TrustRadius, and category-defining trade publications. Brands with G2/Capterra/Trustpilot profiles are roughly 3× more likely to be cited in AI answers for product queries (Digital Bloom). In consumer ecommerce, the graph is wider: Reddit threads, YouTube reviews, and category sites like Wirecutter or NerdWallet do disproportionate work. In financial services and healthcare, AI models lean hard on government and standards bodies plus a small set of named publications, so a feature in the right vertical outweighs a dozen general-business mentions. The corollary: the same PR retainer that funded ten generic tech-blog placements last year should fund three to five carefully chosen tier-one or category-defining placements with body-text mentions and original data.

How to measure earned-media-to-citation lift

The measurement model the PR industry inherited tracks impressions, links, and share of voice across owned monitoring tools. None of those track AI citations. The relevant unit is "named citations per priority prompt, by source," measured weekly across the AI surfaces. The mechanics: define a 25–50 prompt set per priority category, run it weekly in ChatGPT, AI Overviews, and Perplexity, log every citation with parent domain and named brand presence, and attribute lift to the earned-media placements that landed in the same window. Build the baseline before the PR campaign starts; the lift is the change. For the prompt-set discipline, see how to build an AI visibility prompt set, and for the gap-finding workflow, AI citation gap analysis. The most common surprise for teams that run this measurement honestly is that placements they were proud of did not move citations and placements they undervalued did.

The 67% that you cannot pitch directly

Two thirds of ChatGPT's top 1,000 citation sources are domains marketers cannot directly influence: Wikipedia, Reddit, government sites, academic and standards bodies (Ahrefs). That fact gets misread as discouraging. The right read is that earned media is the indirect lever. Wikipedia editors source new entries from named coverage in tier-one publications, so a news placement two quarters ago is often the prerequisite for the Wikipedia entry that anchors entity authority today. Reddit and Quora threads cite vendor pages and reviews; an earned placement on a high-trust review site or a comparison page often shows up in community discussions weeks later, which the AI then cites. The 67% is downstream of the 33%. The PR plan that ignores the indirect path optimizes for short-cycle hits and underperforms on the cycle that compounds.

What this means for the PR budget conversation

The honest reframe for the conversation with leadership: the existing PR retainer is now also the AI visibility line item, and the metrics should match. Stop reporting placements as the deliverable and start reporting citation lift against a priority prompt set and a named-source pitch list. Reduce activity on outlets that produce neither tier-one entity coverage nor AI citation lift. Reallocate the savings into a slower cadence of high-trust placements that earn body-text mentions with structured supporting data. Pair the PR program with a quarterly audit of source citations against competitors, surfaced in Parse's Citations view at /sources, and the budget conversation moves from "we got X mentions" to "we closed Y citation gaps and the prompts that map to revenue now name us." For the broader context on how PR fits into AI strategy, see ghost citations: brand strategy and community visibility: the AI citation pipeline. If you want a team to execute this for you, Soar handles managed AI visibility.

FAQ