Earlier this week I attended Mastering GEO for Comms Pros, a half-day invite-only event hosted by Amanda Coffee and Noah Greenberg of Stacker in an overcast Midtown Manhattan conference room. The room was full of senior communications leaders, and the conversation was focused on one question: how does earned media perform in an AI-driven search world?
The answer, it turns out, is that it performs extraordinarily well - but only if you know what you're optimizing for. Here is what stood out.
Earned Media Is the Currency of AI Search
The statistic I kept replaying: 94% of all AI citations come from non-paid media, with 82% of those coming from earned coverage. For a room full of communications professionals, that number reframes the entire value proposition of what we do. Every placement we secure is not just reaching a journalist's audience but it is potentially shaping the answers that AI surfaces to millions of people daily. The implications for how we position our work to clients shouldn't be understated.
Consistency Trumps Talent
One of the clearest points of alignment across every panel: LLMs reward recency, uniqueness, and repetition. A single strong placement no longer moves the needle the way it once did. Data shared at the event showed that content visibility typically peaks around day 20 and sustains for only five to seven days before dropping sharply. Volume and cadence matter as much as the quality of any individual hit, reinforcing that communications professionals need to be thinking in campaigns versus moments.
Writing for LLM Ingestion and What It Actually Entails
This is where the conversation got most practical, and where I think our team has the most to gain. One of the most useful conversations I had was with the VP of Brand Strategy at Axios, after her panel. She explained that Axios' signature format - short paragraphs, bolded leads, clear hierarchy - is not just a reader experience choice but rather is optimized for LLM ingestion. That discipline is built into how every Axios reporter is trained from day one, and it is entirely purposeful.
The reason it works is largely structural. LLMs parse and cite content that is easy to extract and contextualize. Dense, meandering paragraphs get deprioritized - data used to train it was straightforward. Clearly labeled sections, direct declarative sentences, and explicit transitions between ideas all make it easier for a model to identify what a piece of content is actually saying - and to surface it in response to a relevant query. Newsrooms globally have figured this out early while I’ve witnessed entire communications teams that have not yet caught up.
The tradeoff worth naming honestly: writing for LLM ingestion can feel a bit humiliating at first, arguably reductive at the least. There is an ever-present tension between the kind of nuanced, narrative-driven storytelling that makes a great feature and the structured, signal-rich format that AI rewards. The two are not mutually exclusive, but they require intentionality. A well-crafted story with no structural logic is harder to cite. A rigidly structured piece with no narrative pull won't get placed. The goal is both.
For our team, the discipline starts internally. How we write recaps, notes, action items, and client-facing documents is increasingly the same behavior loop that shapes how information gets consumed and reinforced at scale. The sooner we internalize that, the easier it becomes to counsel clients on the same. Colleen shared a phrase that stuck with me: be worthy of someone's time. It applies to pitching just as much as publishing. Less spray and pray. More precision. Understanding how a reporter or editor thinks - seeing around corners, speaking their native language - is what separates coverage that lands from coverage that gets ignored.
That has always been true. But in a world where AI is the new first impression for brands, getting it right matters more than ever.