Why PR is the new SEO: earned media's role in AI search

Throughout my career, I've watched search evolve and told clients and stakeholders the same thing every time: coverage is currency. What's changed is who's spending it. It used to be Google. Now it's ChatGPT, Perplexity, Gemini and Google's own AI Overviews, and they're all reading the same sources that both Brand Building PR and Digital PR have always chased. That's why the discipline everyone was quietly deprioritising in favour of paid and performance channels is suddenly the most valuable lever a brand has.

Marketers have started calling the work of showing up well in these answers Generative Engine Optimisation, or GEO.

What GEO actually means

At its core, GEO is the practice of shaping how AI systems describe, recommend and cite a brand when someone asks a question rather than types a keyword. There's no meta tag for this. Large language models build their answers from a blend of training data and live retrieval, and what they retrieve is overwhelmingly earned media: news articles, trade press, expert commentary, forum threads, review sites, Wikipedia. Brand-owned content matters far less here than most marketing teams assume, because AI models are trained to distrust anything that reads like self-promotion. Third-party validation is the raw material these systems are built to trust.

That's the entire argument for PR in one sentence. If AI answers are assembled from what other credible sources say about you, then earned media is no longer a nice to have for reputation. It's the input layer for how you show up in AI search. This is the same thinking behind Brand Building Through Earned Media, and it's worth reading if you want the fuller strategic case.

Why brands are getting this wrong

Most marketing budgets are still allocated as if 2019 never ended: heavy on paid social, heavy on traditional SEO content, and PR treated as a reputation insurance policy you buy after a crisis rather than a growth channel you invest in before one. I see this constantly in pitches and case studies where a brand has poured money into a beautifully optimised website that no AI model ever cites, because nothing outside that website talks about the brand in a way worth referencing.

The mistake is treating AI visibility as a technical SEO problem when it's fundamentally a trust and distribution problem. You can't structure data your way into a ChatGPT answer. You get there by being the source journalists quote, the case study trade press covers, and the brand independent experts mention unprompted. That's Digital PR Strategy in a nutshell. It always has been. The channel it now feeds has just changed.

The four layers of earned media that feed AI visibility

I look at AI visibility against five conditions: authority, consistency, corroboration, recency and citation (I will cover this in more detail in another The Spark post coming soon).

Block one below, foundational facts, builds the baseline authority and consistency every earned media campaign depends on, and I think this should be handled by your content and finance teams, collaborating to ensure the latest and most accurate information is used.

The three genuinely earned media blocks that follow, trade press, national coverage and expert commentary, are what build corroboration, recency and citation.

Here's how that work breaks down in practice:

Block one: foundational facts. Wikipedia, industry directories, company registries. Dull, unglamorous, and the first thing most AI models check for basic factual grounding. If your Wikipedia page is thin or your Companies House details don't match your website, you're starting from a deficit before a single press release goes out.

Block two: trade and hyper-relevant press. The publications your industry actually reads. These help AI models associate your brand with a specific category or topic, which matters for whether you get mentioned at all when someone asks about that category. That's a different kind of value to a single national splash, not necessarily a bigger one. Citation studies consistently show major national and reference sources, Wikipedia, Reuters, the New York Times, Forbes, dominate raw citation volume, so trade press is best understood as building topical association rather than outperforming national coverage on citation count.

Block three: mainstream and national coverage. This is where most PR budgets still concentrate, and it matters, but its AI value comes less from the outlet's authority and more from how often the story gets syndicated and referenced elsewhere. A story that only lives on one masthead has a shorter half-life in AI training and retrieval than one that gets picked up, quoted, and linked across the wider web.

If you've worked with me before, you'll know what a shift this is. In the old world of digital PR, built to satisfy the Google gods, a syndicated link had arguably low value, a duplicate, canonicalised away, worth little next to an original placement. AI has changed that for me. Language models are looking for consistency and corroboration across the web, not a single canonical source, so a story picked up and repeated across ten sites now carries more weight than one that only ever lived on the original masthead. It's a genuine 180 on something I spent years telling everyone to deprioritise, and it fits the data: Ahrefs' research found unlinked brand mentions correlate more strongly with AI visibility than backlinks do, which is exactly the kind of signal syndicated coverage generates even when no single copy carries the canonical link.

Block four: earned commentary and expert positioning. Founder and spokesperson quotes, expert roundups, contributed opinion pieces, forum and community mentions. This is the layer that teaches AI models to associate a named person with a subject, which is exactly how these systems build the who should I ask about X?mental model behind their answers.

Does it matter if the site mentioning me is relevant to my industry? In short, yes. This is the drum I've been banging for years, and AI hasn't changed that. Being mentioned somewhere authoritative is not the same as being mentioned somewhere relevant. A home insurance brand covered widely on general news sites can still be invisible in AI answers about home insurance, because nothing in that coverage ties the brand to the category. A brand mentioned less often, but consistently in home insurance content, builds a clearer association for the exact questions it wants to be cited for.

Authority gets a story picked up. Relevance decides what that story is actually remembered for.

What this looks like in practice

This is the thinking behind the hyper-relevant statistics pages I built across the RVU portfolio, including:

• Broadband statistics for Uswitch

• Temporary car insurance statistics for Confused.com

• Small business statistics for money.co.uk

The money.co.uk page alone has earned more than 550 backlinks over four years, ranks second in Google for "business statistics" ahead of the ONS and GOV.UK, and has been cited directly in a Google AI Overview, itself a small but concrete proof point for the argument this whole post is making. You can read more about this on my Work page.

Each page was built to be hyper-relevant to its category, not a broad overview trying to rank for everything, but content specific enough to rank for the exact terms its audience searched, and specific enough for other sites writing about that exact topic to reference it. That's relevance compounding in both directions: relevant content ranks for relevant keywords, and earns pickup from relevant, on-topic sources. It's precisely the kind of tightly scoped, category-matched footprint AI retrieval rewards over a single broad page trying to cover everything at once.

The practical shift for most comms teams is threefold.

First, stop measuring PR success purely on reach and start measuring citation quality: ensure your brand is described accurately and consistently across the sources that matter, in the same language, with the same facts.

Second, invest in the boring block one work. Audit your Wikipedia presence, your directory listings, your structured company data. It's unglamorous, and it's foundational.

Third, treat expert positioning as a distribution strategy, not a personal branding nicety. Every spokesperson quote that lands in a trade feature or expert roundup is training data for how AI systems answer questions in your category. If you're building this out, it sits close to the ground covered in Influencer Marketing Strategy, since third-party voices work the same way whether they're journalists or creators.

The opportunity

Here's what makes this exciting rather than just another channel to manage: most competitors haven't worked this out yet. Budgets are still weighted towards channels built for a search era that's already ending. Brands that shift real investment into structured, sustained earned media now will own the AI answer for their category before their competitors even notice the question has changed. That's a rare position to be in in comms: not catching up to a shift, but getting ahead of one while it's still forming.

PR was never just about column inches. It was always about being the trusted source other people repeat. That's the exact mechanism AI search runs on. The discipline hasn't changed. The stakes just got a lot higher.

If you are looking for senior communications expertise that combines strategic instinct with smart use of AI, get in touch to discuss how I can help.