PR Just Became a Growth Metric: Proving the Link Between Media Placements and AI Visibility

For decades, PR’s business case rested on soft metrics and good faith. That’s no longer good enough and it’s no longer necessary.

Every comms leader has sat across from a CFO and struggled to answer the same question: what did that coverage actually do for the business? For years the honest answer was “it built awareness, credibility, trust”. Real, but hard to put a number against. That’s why PR budgets have always been the easiest line item to question and the hardest to defend.

AI search has changed the terms of that argument, in PR’s favour. The data now says, plainly, that earned media is the single biggest driver of whether a brand shows up when someone asks ChatGPT, Claude, Gemini, or Perplexity a category question. The challenge has shifted from “does PR matter?” to “can we prove it, in a way leadership will believe?”

Why PR Now Sits at the Centre of AI Visibility

The evidence here isn’t a single vendor’s pitch deck, it’s converging findings from multiple independent studies, run by different companies with different methodologies, all landing in the same place. I first picked up on the criticality of PR for AI search at a conference in 2025, it was like a lightbulb going off in head, it made sense from the get go.

Muck Rack’s What Is AI Reading? research, now three editions deep and covering more than 25 million citations across ChatGPT, Claude, and Gemini, has found earned media accounts for 82-89% of all AI citations, consistently, since mid-2025. This is the proof we’ve all needed despite our guts telling us this was the way to go. 

Ahrefs studied 75,000 brands and found brand web mentions (the classic PR outcome) correlate with AI Overview visibility three times more strongly than backlinks, the metric SEO has optimised for over the past two decades. Brands in the top quartile for web mentions earned ten times more AI Overview visibility than the next quartile down. Brands in the bottom half were essentially invisible.

Perhaps the most useful data point for anyone trying to make this case internally: Stacker and Scrunch ran a controlled study distributing the same content through third-party news outlets versus keeping it on a brand’s own site. Citation rates went from 8% to 34% – a 325% lift – purely as a function of where the content lived. Not what it said but ‘where’ it appeared.

McKinsey’s read on this is the plainest of all: a brand’s own website now accounts for only 5-10% of what AI search references when constructing an answer. The other 90-95% is third-party – media coverage, review sites, forums, affiliate content. If your AI visibility strategy is built entirely around your own site, you’ve optimised for the smaller channel and largely ignored the one that actually decides whether you’re in the conversation.

Put simply: the mechanics that have always made PR valuable – third-party validation, editorial credibility, a journalist vouching for you rather than you vouching for yourself – are precisely the signal AI models were trained to trust. A brand talking about itself is recognised as self-interested. A respected outlet talking about that brand is not. That distinction is baked into the models, which is why no amount of on-site optimisation substitutes for earned coverage.

The Problem: This Is Genuinely Hard to Prove

None of this means the business case writes itself. Showing the connection between media activity and AI visibility is a different, harder problem than showing that the connection exists in the abstract. It’s like the Holy Grail of ‘proof’, and PR has always sat at the heart of this particular conundrum. 

There’s no click to point to. 

Traditional digital marketing built its credibility on last-touch attribution: a visitor clicked this ad, landed on this page, converted. AI answers don’t work that way. A prospect asks an AI model who the category leaders are, forms an impression, and may never generate a trackable event you can tie back to a specific placement.

The effect compounds slowly, not instantly. 

A placement in a trusted publication doesn’t shift AI citation behaviour overnight. Models index, cross-reference, and re-weight authority over weeks and months, not days. That lag makes it easy for a skeptical stakeholder to look at a campaign and conclude nothing happened, when the effect simply hasn’t surfaced yet.

Every platform behaves differently. 

ChatGPT cites in 96% of responses but averages five sources. Claude cites in only 55% of responses but averages thirteen when it does, and leans toward long-form editorial. Gemini leans on Reddit. A strategy that lifts your visibility on one platform may do nothing on another – which makes “did it work” a genuinely multi-part question rather than a single number.

The models are a black box. 

You cannot see what a given AI system was trained on, when its knowledge was last refreshed, or exactly how it’s weighting your latest placement against everything else it has read about your category. You’re inferring cause from correlation, not measuring it directly. And for those of who have spent years keeping on top of the Google algorithm for SEO, we now have to monitor multiple AI models to ensure we don’t miss out. 

Leadership has heard this story before. 

PR has spent years being asked to prove ROI against metrics: impressions, AVE, sentiment. And we all know none of that ever quite held up under scrutiny. Anyone bringing a new claim about AI visibility to the board is going to be met with fair, earned skepticism. The case has to be built on numbers, not narrative.

How to Actually Show the Connection

The good news is that this is now measurable … not perfectly, but well enough to build a credible, defensible case.

Establish an AI share of voice baseline before you do anything else. 

Platforms like Profound, Scrunch AI, Peec AI, and Ahrefs’ Brand Radar exist specifically to track how often your brand, versus named competitors, gets surfaced across a defined set of category prompts on ChatGPT, Claude, Perplexity, and Gemini. Run this for four to six weeks before any campaign activity so you have a genuine “before” to compare against, not a guess. Use the same tool, use the same criteria, use the same prompts and be a bit insular, you are comparing evolving results against yourself not against a benchmark. One tiny adjustment can throw off the score so consistency is key. 

Track placements and AI share of voice on the same timeline, not in separate reports. 

Most PR reporting and most AI visibility reporting currently sit in entirely different documents, built by different teams, on different schedules. Put media placement volume and tier (national press versus trade versus podcast versus press release) on the same monthly view as AI share of voice movement. Correlation isn’t causation, but a consistent pattern – placements cluster, AI visibility rises two to eight weeks later, repeatedly, across multiple cycles – is a genuinely strong argument, and it’s the same standard of evidence the Ahrefs and Stacker studies used to make their case publicly.

Borrow the controlled-test logic from the Stacker/Scrunch study. 

You don’t need five LLMs and 944 prompt combinations to run a smaller version of the same idea. Take a piece of content or a story you already have. Place it in a trusted third-party outlet. Track AI citation for that specific narrative before and after the placement runs, isolated as much as possible from other activity in the same period. A single clean before/after comparison, repeated a few times, builds a portfolio of evidence that’s far more persuasive than a single industry statistic.

Report citation, not just mention. 

Being mentioned by an AI model and being cited as a source it links back to are different outcomes, and the more valuable one – the one tied to referral traffic and consideration – is citation. Track which specific placements are actually showing up as cited sources, not just contributing to a general sense of brand presence.

Segment by platform, and set expectations accordingly. 

Report AI share of voice per platform, not as a single blended number. If Claude favours long-form editorial and Gemini favours Reddit, a placement strategy has different expected payoffs on each, and reporting them separately protects you from someone concluding “this doesn’t work” when it worked exactly where the mechanics said it would.

Frame the metric in language leadership already trusts. 

AI share of voice is a stronger case internally when it’s presented as what it structurally is: a leading indicator for a channel that 94% of B2B buyers are already using in their purchasing process, per Forrester’s most recent data. This isn’t a vanity metric competing with pipeline and revenue numbers – it’s an early signal of whether your brand is even in the consideration set before those numbers get generated.

In Conclusion

The evidence that PR and media placements drive AI visibility is no longer in question, it’s been replicated across studies from PR analytics firms, SEO data companies, and academic researchers, using different methods and reaching the same conclusion. What’s still genuinely hard is showing the connection convincingly, campaign by campaign, to people who’ve been burned by soft PR metrics before.

That’s solvable, but it requires treating AI share of voice the way performance marketing treats any other funnel metric: baseline it, track it on the same timeline as the activity meant to move it, and report it consistently enough that the pattern becomes obvious rather than asserted. The brands doing this now aren’t just ahead on measurement, they’re building the case that will make PR budgets easier to defend for years to come, at exactly the moment AI is becoming how a growing share of buyers form their first impression of a category.