AEO for PR: How to Prove Your Earned Media Actually Moved the AI Answer

Questions this guide answers

• Does PR or earned media affect AI search visibility?
• How do I measure the impact of a PR campaign on ChatGPT and Perplexity?
• What should a PR team report about AI visibility?
• Why is a PR hit easier to measure in AI search than content?
• Which earned media moves AI answers the most?

Direct answer

For a PR or comms team, the measurement question that matters is no longer "did we get coverage," it's "did the coverage change what AI says about us." That is measurable, and it's more measurable than most content, because a media hit is a discrete, dateable event: you get a clean before and a clean after.

Earned media is arguably the highest-leverage AEO work a brand can do. AI decides what to recommend by corroborating across the sources it trusts, and it weights independent third-party coverage higher than your own site precisely because it isn't you talking about yourself (see AI cites consensus, not authority). That is the definition of earned media. The problem was never leverage, it's that PR measurement was built for impressions and reach, not for "did the model change its mind." Here's how to measure the thing that now matters.

Why earned media is your strongest AEO lever

Start with the reframe, because it changes what PR is worth. When a buyer asks an assistant "who's best at X," the model doesn't read your website and repeat it. It assembles an answer from the sources it trusts, and the ones it trusts most are the independent third parties, because they aren't the brand talking about itself. A trade outlet, an analyst note, a respected journalist saying something about you is exactly the kind of evidence that moves an AI answer, in a way your own "why we're the best" page never will, the model discounts that page as a self-claim.

So PR isn't adjacent to AEO. Earned coverage IS the corroboration layer, which makes it close to the core of how AI recommendations are formed (see owned, earned, and community sources). The uncomfortable flip side is that a great campaign that doesn't change what the model says didn't fully land, and until recently PR had no way to see that either way.

Why a PR hit is more measurable than content

Most AEO measurement struggles with attribution, because content is diffuse. You publish a page and it slowly, maybe, helps, with no clean line between the change and the result. A media hit is the opposite: a specific piece of coverage with a date. That hands you the cleanest natural experiment AEO offers, a clear before, a clear intervention, and a clear after, timestamped.

That's the opportunity comms teams are sitting on and not using. The event PR is organized around, the hit, the announcement, the campaign launch, is precisely the kind of dated intervention you can run a real before/after against. The method for doing that honestly is the same one that keeps any AEO experiment from lying to you (see the testable GEO playbook); the rest of this page applies it to a PR event.

The framework: measuring a hit

Five steps turn a media hit into a measurable result instead of a slide of impressions.

• Freeze the buyer prompts the coverage is relevant to. Not vanity prompts, the questions your buyers actually ask where you'd want the model to now mention you or say the thing the coverage said. If the feature was about your speed, the prompts are the ones where speed decides the pick.
• Baseline before the hit. Run that set for a couple of weeks before the coverage lands, to get your share of voice and a read on how much the answers move on their own. One pre-hit run is not a baseline, it's a single noisy draw.
• Watch the specific claims, not a single score. The causal fingerprint isn't "our visibility number went up," it's "the specific thing the coverage said started appearing in the answers." If a trade outlet called you the fastest to deploy, watch whether "fast" enters how the model describes you. That's the hit working.
• Read it as a rate, and expect a lag. The coverage has to be crawled and indexed before it can move an answer, so weeks one and two may show nothing even in the success case. And the answers are noisy, so a single post-hit run tells you little. Judge it over weeks.
• Measure the queries the coverage addressed, not your overall visibility. Earned media moves the specific ground it covered, not your whole footprint. Averaging across everything dilutes the signal until it disappears.

What actually shows the AI impact

The metrics PR reports today mostly can't see the AI effect. Here's the swap.

What to report to a PR stakeholder

Don't hand a CMO or a client a single AI-visibility number, they'll rightly distrust it, and they should (see is AI visibility a vanity metric). Report three things: the buyer prompts where your share of voice moved after the hit, the specific claims from the coverage that now show up in AI answers, and the honest caveat that this is a leading indicator with a lag, not a same-week revenue line.

The framing that lands with an executive is simple: the coverage didn't just reach people, it changed what the machine says about us when a buyer asks. That's a sentence a PR team could never say before, and it reframes earned media from an awareness cost into a measurable input to how buyers get recommended.

The honest version

A few caveats keep this credible. Attribution is never perfectly clean, the model saw a great deal that week besides your hit, which is exactly why you watch the specific claims and the specific queries rather than a blended score. The demand for the query has to exist, a hit on a question nobody asks moves nothing you'll be able to see. And not all earned media is equal: the model leans on the outlets it already trusts and corroborates elsewhere, so a placement in a source the model doesn't weight moves less than its raw reach suggests (reading your earned sources by how much the model actually trusts them is its own exercise, see the incentive map).

A worked example

Take a representative case, the comms team at a data-infrastructure vendor we'll call Northwind (not a real company). They landed a trade-press feature that called them the fastest in their category to deploy. The old report would have been the 40,000 impressions.

Instead, two weeks before the feature ran, they'd frozen 25 buyer prompts, including several where deployment speed decides the pick, and baselined them. Pre-hit, the engines never mentioned speed for Northwind. About three weeks after the feature, allowing for the index lag, "quick to deploy" and "fast setup" started appearing in how ChatGPT and Perplexity described them, but only on the queries the feature actually addressed, their overall visibility barely moved. The report to the CMO wasn't "we got coverage." It was "the model now tells buyers we're fast to deploy on the questions where that matters, and here's the before and after."

See your pre-campaign baseline

The measurement only works if you capture the before. Start by seeing what the engines say about you on your buyer prompts today, that's your baseline for the next hit. Start Free (free, no credit card) and SolCrys shows you where the five major engines mention you, which sources they cite, and how they describe you, so you have a real "before" to measure a campaign against.

Talk to us if you want it run continuously through campaigns, tracking the specific claims and the specific prompts a hit was meant to move.

Earned media was always about a trusted third party changing what an audience believes about you. The audience is just the model now, and for the first time you can measure whether it worked.

FAQ

Related guides