Every content team obsesses over rank. Most AI citations do not come from the page ranking first. They come from a journalist's article, a comparison blog, or a forum thread that a search team never touched, and that page usually isn't in your top 10 either.
Rank isn't the gate you think it is
ConvertMate's 2026 GEO benchmark found 83% of AI Overview citations come from pages sitting outside Google's organic top 10 for the query. That number alone should change how a content team allocates effort. Chasing position one on a keyword and chasing an AI citation on the same query are, increasingly, two different jobs with two different playbooks.
Where earned media AI citations actually come from
| SOURCE TYPE | SHARE OF AI CITATIONS |
|---|---|
| Earned media (editorial coverage) | 82% |
| Paid or owned content | 6% |
| Journalism specifically | 20-30% |
Muck Rack's December 2025 research put earned media, coverage you did not pay for and do not own, at 82% of all AI citations tracked. Paid and owned content combined accounted for 6%. Journalism specifically, a subset of that earned-media share, accounted for 20 to 30% on its own. If your GEO plan is entirely about publishing more owned pages, it is optimizing 6% of the pie, the same imbalance we flag in the six link types AI engines actually trust.
The same research found press release citations climbing 5x between July and December 2025. That is a separate, growing channel: AI engines increasingly cite the release document itself as a standalone source, whether or not a journalist ever picked it up.
Why earned media AI citations beat owned content
Third-party citation sources, for comparison (Kevin Indig's agent study, Jul 2026)
The pattern holds even outside Muck Rack's dataset. When AI agents can't find a first-party answer and fall back to a third party, editorial coverage and directories split almost the entire fallback between them, not your owned pages. Earned media isn't one lever among many for AI visibility. Across two independently run studies now, it's most of the mechanism.
This is also why link building for GEO looks less like acquiring backlinks and more like our approach to link building for the AI era: the goal isn't a domain-authority number, it's a mention from a source an engine already trusts enough to cite on its own.
It's also worth pointing at what that citation is actually worth once it lands. Adobe Digital Insights found AI referral traffic grew 693% year over year during the 2025 holiday season, with AI-referred conversions outperforming non-AI traffic by 31%. Earned-media citations aren't a brand-awareness nice-to-have anymore; they're feeding one of the fastest-growing, higher-converting traffic sources most teams have, which is exactly why the 2% overlap number should worry a CMO as much as it worries the PR team.
Community platforms belong in the same target list as journalists. A June 2025 Semrush study found Quora the single most-cited source across Google's AI Overviews, with Reddit ranking second, ahead of most traditional publishers. Neither platform runs on pitches. They run on genuine, specific, useful answers left in the threads where your buyers are already asking the question. A rebuilt target list that only adds bylines and skips the communities those bylines actually read is only half the fix.
It's worth sitting with why the overlap is this low in the first place, rather than treating it as a random mismatch. Traditional PR targeting optimizes for outlet prestige and audience size, the metrics that mattered when the goal was a human reader clicking through from a homepage. AI citation targeting optimizes for something almost unrelated: whether an engine's retrieval system already trusts a domain enough to lift a claim from it with confidence. A trade publication with a small readership but a long history of being crawled, indexed, and cited accurately can outperform a top-tier outlet an engine has less reason to trust on a specific technical claim. The overlap is low because the two targeting systems were built to optimize for different outcomes, not because PR teams are bad at their jobs.
The walkthrough: rebuild your target list
Step 2 deserves a specific warning, because it's where most teams flinch. Seeing a 2% overlap number against your own list feels like an indictment of years of relationship-building work, and the instinct is to explain it away: maybe the query set was too narrow, maybe the sample was unlucky. Run the check on more than ten queries before you trust that instinct. In our own audits across client media lists, the number rarely moves by more than a few points once the query set gets to fifteen or twenty prompts. The overlap really is that low, for almost everyone, and the sooner a team accepts that, the sooner the rebuild in steps 3 and 4 actually starts.
One more reason to run this audit before your next campaign, not after: the same Muck Rack research found press release citations themselves climbing 5x in six months, which means the standalone release is quietly becoming a legitimate citation asset independent of whether a journalist ever covers it. A rebuilt target list and a well-written release are not competing tactics. Run both, and measure them separately, since they're earning citations through different mechanisms.
What changes for how PR gets measured
This also changes what a PR team should report to leadership. Placement count and estimated reach were always imperfect proxies for value, but they were at least directionally tied to human eyeballs. Once 82% of AI citations come from earned media, a placement's value increasingly runs through a second channel entirely: whether that specific article becomes a source an AI engine leans on repeatedly across dozens or hundreds of buyer prompts, long after the article's original traffic spike has faded. A single well-placed, well-sourced piece of coverage can keep paying out in AI citations for months after a press release's news cycle ends, which is a return on investment traditional PR reporting was never built to capture.
That kind of return is exactly why this walkthrough is worth running as a recurring quarter-over-quarter process, not a one-off audit. Journalists move outlets, new publications earn AI trust, and old ones lose it. A target list rebuilt once and never revisited will drift back toward the same 2% overlap within a year.
Do this next
Run step one this week, on your own top queries, before you plan another press release. The list you get back will not look like your media database. That mismatch is the whole finding, and it is the cheapest audit your PR function will run all quarter. It's the same first move we make on our digital PR and link building engagements, and it's how a recent B2B staffing client found the three trade outlets actually driving their category's AI answers, none of which were on the original media list.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Josh leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.