Performance PR Has a New Definition. and Most Agencies Don’t Know It Yet

There is a term in my industry that has meant the same thing for 15 years. Performance PR. It sounds rigorous. It sounds like accountability. For most of my career, it described a simple deal: the agency only gets paid when the article actually publishes. No placement, no fee. I built my company on that model because it forced honesty. You delivered, or you did not eat.

Here is what almost nobody in the industry wants to admit. That definition of performance now measures the wrong thing. An article can be published, the fee can be earned, the report can look flawless, and the placement can still do nothing for the brand that paid for it. The deal performed. The PR did not.

Let me be clear about what I am not saying. PR is not dead. Earned media is more valuable right now than at any point I have watched this industry, and I have watched it for the better part of a decade. A real placement in a real publication still builds credibility, still opens the door to an investor, still gives a founder something true to stand on. None of that has weakened. What broke is the meaning of the word performance, and almost no firm has noticed.

Performance used to be a question about activity. Did the article go live? Did it hit the agreed numbers, the impressions, the domain authority, the reach? Those were answerable, and for a human audience, they were enough. But the first reader of that coverage is no longer a person scanning a homepage. It is an inference engine deciding whether your brand is worth repeating when a buyer asks the exact question your company exists to answer. Activity is not the same as being repeated. A brand can earn 50 placements in a quarter and be absent from every answer the machines give. The agency performed. The brand vanished.

Put simply, performance used to mean the article published. Now, in the AI search era, performance PR means the machines repeat it. All of them. Because there is no single machine to win anymore.

A study of 17.2 million AI citations found that each major model pulls from a different set of sources. A brand that owns the answer inside ChatGPT can be invisible inside Perplexity. One that dominates Google’s AI Mode might not exist in Claude. Your buyers are not loyal to one engine. They use whichever one they happen to trust that morning.

So performance, if the word is going to mean anything, cannot mean winning in one place. It has to mean corroboration. The same brand, making the same verifiable claim, showing up across enough independent sources that every model arrives at the same answer about you. Gartner’s 2026 research points the same way: as AI-written copy floods every channel, machines increasingly reward brands with specific, corroborated and verifiable claims over the ones producing the most words. Volume was an advantage in PR for 30 years. Now it is noise.

This is also why the placement still matters more than your own website. 89% of the citations AI engines make come from earned media, coverage on domains you do not own, not from the pages you control. Optimizing your own site helps. Worth doing. But the machines trust what other people publish about you far more than what you publish about yourself. Earned media is the infrastructure. Corroboration is the product.

The founders who see this early get something that compounds. When your earned coverage is built to be corroborated across every model a buyer might open, your brand becomes the default answer in your category. Not one option in a list of 100. But the answer ChatGPT gives. The answer Perplexity gives. The answer Google’s AI Mode gives. The answer Claude gives. That is not a campaign you run and refresh every quarter. It is a position that gets harder for a competitor to take with every consistent placement you add.

Inside AuthorityTech, I rebuilt the entire performance-based earned media practice around that single bet, that the audience deciding now is a machine and the win is being corroborated by all of them. I named the discipline Machine Relations in 2024 because nothing else described both the shift and the work it demands.

So here is what I tell founders when they ask what this means for the PR budget they are already spending. It does not require spending more. It requires changing what you call success before you sign the next contract.

Redefine performance before you sign anything

Most contracts still define a win as publication. The article goes live, the invoice clears and everyone moves on. Change the definition in writing. A placement performs when an AI engine repeats it as the answer to a real buyer question, not when it shows up in a media report. If your agency cannot explain the difference, they are selling you the 2010 version of the word.

Test your coverage in every model, not your favorite one

Open ChatGPT, Perplexity, Google’s AI Mode and Claude. Ask each one the five questions your ideal customer asks before buying from anyone in your category. Do not search your brand name. Search the problem. Write down where you appear and, more importantly, where you disappear. Most founders find they exist in one engine and vanish in the other three. That gap is your real performance score. If you would rather see it scored for you, free audits can run the same check across the major models, one inside ChatGPT and one inside Gemini.

Make one claim survive across every article

Corroboration only works if the machines keep seeing the same thing. If one placement says you are the fastest, another says the most affordable and a third says the most secure, you have handed the models three different brands to choose from. Pick the single hardest, most specific claim you can defend. Use the sharpest number you have. Then make every piece of coverage carry it. Consistency is what turns separate articles into one corroborated answer.

Pay for outcomes the machine can see

If you use a performance model, and you should, tie the fee to something the machines actually register. Publication is not it. Citation is. The recommendation is. An agency confident in its work will agree to be measured on whether the coverage shows up in the answers, because that is the only performance that moves a pipeline now.

The PR firms that survive this are not the ones placing the most articles. They are the ones who understood that the word performance changed meaning and rebuilt around the new one. The rest will keep sending impression reports that describe a kind of success the machines stopped counting years ago.

So before you renew, ask the only question that matters now. Not just whether your firm can get you placed. Whether anything it places will still be repeated, in every model your buyer trusts, six months from today. The founders who answer that honestly are the ones the machines will keep recommending. The ones who do not will wonder why their best quarter of coverage bought them nothing.