How PR Earns Your Brand a Citation in ChatGPT and Google AI Overviews

AI search tools do not cite adverts, and they rarely cite a brand's own website above everyone else's. They cite earned, third-party coverage, the same reporting and expert commentary PR has always existed to secure. If a brand wants ChatGPT, Google AI Overviews or Perplexity to describe it accurately when someone asks a relevant question, the fastest route is not a better meta tag. It is a strategic, ongoing media relations programme that produces the kind of coverage these systems are built to trust.

Why do AI search tools prefer earned media over your own website?

Large language models are trained to weight third-party validation heavily, because a brand describing itself is a far weaker signal than an independent outlet describing it. Earned coverage in a recognised publication carries an implicit endorsement that owned content cannot replicate, however well it is written. That is precisely the function PR has always performed for search engines and journalists alike. Only the audience reading the output has changed, from a person scanning search results to a model synthesising an answer.

What does the research actually say about what gets cited?

The most cited research on this is the “GEO: Generative Engine Optimization” paper out of Princeton, Georgia Tech and the Allen Institute for AI, which ran 10,000 queries through generative search systems and tested nine specific content strategies against how often they got cited. Adding a direct quotation from a named source improved citation rates by 41%. Adding a specific statistic improved them by 32%. Citing an external source within the content improved them by 30%, and tightening the writing for clarity added a further 28%. In other words, the same things that make a piece of journalism credible, a named quote, a hard number, a cited source, are measurably the things that make it more likely to be surfaced by an AI system.

A separate, large-scale study by CXL, analysing roughly 100 pages of Google AI Overview citation data, found that 96% of AI Overview citations come from sources with strong E-E-A-T signals, experience, expertise, authoritativeness and trustworthiness, and that 55% of citations are pulled from the top 30% of a page. Google's own developer documentation is notably understated about this: its official guidance states there is no special markup or technical trick required to appear in AI Overviews, and that the existing fundamentals of good, crawlable, well-structured content are what matter. The research and the platform's own position agree: there is no shortcut, only the same credibility signals PR has always been built to earn.

Does this replace SEO, or sit alongside it?

It sits alongside it. Technical SEO and structured data determine whether a page can be found and read by AI crawlers in the first place. PR determines whether, once that content exists, an independent source is willing to validate it. A brand with excellent technical SEO and no earned media has given AI systems nothing external to trust. A brand with strong media relationships and a technically invisible website has made itself hard to find at all. The brands winning AI-era visibility are treating the two as one connected discipline, not two separate teams working from different briefs.

What makes a specific piece of coverage citable rather than forgettable?

Citable coverage tends to share a few features consistently. It resolves a specific question clearly, in plain language, without requiring the reader to infer the answer from context. It includes verifiable facts and, ideally, an attributed expert quote rather than only unattributed claims. And it is current, since AI systems favour content published or updated recently over older material making the same point. None of this changes what good PR has always tried to do: get a credible outlet to say something true and specific about a brand. It raises the value of doing that consistently rather than in one-off bursts.

How is earning an AI citation different from traditional link-building?

Traditional link-building optimises for a hyperlink pointing back to your site, often pursued as an end in itself, sometimes through guest posts or directory listings that add little real credibility. Generative engine optimisation, per the Princeton research above, rewards the substance inside the coverage itself, the quotation, the statistic, the named source, whether or not that piece even links back to your website at all. A brand can be cited accurately by an AI system from a piece of coverage that never links to its own site, because the model is drawing on the third-party description of the brand, not the backlink graph. This does not make links worthless, they still matter for traditional search and for driving a reader to click through, but it does mean a PR programme built purely around securing a link, rather than around securing genuine, quotable, fact-based third-party coverage, is optimising for the wrong output if AI visibility is a real goal.

What does a realistic first 90 days of this work look like?

Weeks one to two: map existing third-party coverage against the specific questions your buyers are likely to put to an AI tool, and identify the two or three gaps that matter most. Weeks three to six: secure two or three pieces of genuine earned coverage that include a named quote and a specific, verifiable statistic, prioritising outlets an AI system is likely to treat as credible over ones chosen purely for prestige. Weeks seven to twelve: check technical foundations in parallel, crawlability, structured content, internal linking, and begin testing what a relevant AI search query actually returns for your brand today, so you have a real baseline to measure the next quarter's work against. None of this replaces a longer-term media relations programme. It is simply the fastest, most honest starting point for a brand that has not yet treated AI visibility as a deliberate outcome of its PR.

What mistakes do brands make when they chase AI visibility?

The most common is treating AI citation as a technical SEO task and skipping media relations entirely, adding schema markup and metadata to a website that no independent source has ever written about, which gives AI systems nothing external to validate the claims on the page. The second is a one-off press release blast timed to a launch, then silence for months, when the research above rewards recent, ongoing coverage over a single historical mention. The third is chasing backlinks as a proxy for AI visibility, when the two are related but not the same thing: a link tells a search engine a page exists, while a quotation and a statistic in genuine third-party coverage are what the generative engine research shows actually earns a citation.

How should PR and SEO teams divide this work in practice?

The technical side, crawlability, structured data, page speed, internal linking, sits with SEO, and none of it should be neglected on the assumption that PR alone will carry AI visibility. The earned-media side, securing quotable, fact-backed third-party coverage, sits with PR, and it should be briefed with AI citation as an explicit objective, not an incidental benefit of coverage secured for other reasons. The connecting layer, tracking which pieces of coverage actually get surfaced when a relevant query is put to ChatGPT, Perplexity or Google AI Overviews, works best as a shared, recurring check between both functions, since it tells each side which of their inputs is actually moving the needle and which is not.

What founders and marketing leaders should do first:

• Map where your brand is already being discussed by third parties today, trade press, sector commentary, founder profiles, and identify the gaps against the questions your buyers are actually likely to ask an AI tool.

• Prioritise coverage that includes a specific, attributed quote from a named person at your business, since quotations show the single largest citation lift in the Princeton research.

• Push for data and named outcomes in every piece of coverage you help produce, rather than settling for a mention with no substance behind it.

• Keep technical SEO healthy in parallel, crawlable pages, clean structure, internal linking, since PR without technical access gives AI systems nothing to index in the first place.

• Run media relations continuously rather than in bursts tied only to launches, since recency is a measurable factor in what gets surfaced.

• Treat AI citation as an outcome of running good, senior, targeted media relations, not as a separate technical project bolted onto the marketing team.

Founders who get this right are not chasing AI visibility as a standalone initiative. They are running the same senior, targeted media relations they should be running anyway, and treating AI citation as one more outcome of doing it well.

If you want your brand described accurately when someone asks ChatGPT or Google about it, talk to Fireflies Management.