Before You Trade Your PR Strategy for an AI Content Engine, Read This

I was pitching a prospective client recently when they stopped me with a question that’s been sitting with me for days. They had built an internal system to generate and distribute content across social, web, and product from a single narrative engine. Their plan, as they described it, was to use their “own AI stack as the operating system for distribution” and to “deliberately seed what AI systems see and cite.” They saw PR as a slow lane. Why pitch journalists when you can build infrastructure?

It’s a genuinely interesting idea and one I think a lot of future-forward companies are considering. But the research points in a different direction, and ignoring the role of PR and earned media is a choice with real consequences. Here are five things worth knowing before you cut earned media out of your strategy.

1. AI Models Look for Outside Voices, Not Inside Ones

When someone asks an LLM like ChatGPT or Perplexity about your company, the answer isn’t pulled from your website. It’s drawn from journalists, analysts, and industry publications that have written about you independently.

Muck Rack analyzed over a million links cited in AI responses and found that more than 95% came from non-paid coverage. The models are built to look for external validation. They want multiple independent sources agreeing on who you are before they’ll describe you with any confidence.

Think of it like a reference check. You can write your own resume, but the employer is going to call the people who’ve worked with you.

2. You Can’t Buy a Spot in an AI Answer

Paid media and sponsored content have almost no influence on what these models surface. That’s a big change from how search visibility used to work.

Edelman found that up to 90% of citations driving brand visibility in AI responses come from earned media. There’s no ad buy for an AI answer. The model is looking for credibility signals, and paid placement doesn’t register as one.

For companies willing to do the work, that’s actually good news. The playing field is more level than it was in the pay-to-play search era.

3. Mass-Produced Content Has a Short Shelf Life

Here’s something that gets missed in the excitement around AI-generated content: models are getting better at detecting it, and readers already distrust it. Google’s quality raters actively flag thin, repetitive, or machine-generated content. Publishers are tightening standards. The phrase “AI slop” exists for a reason.

More importantly, volume is not what AI models reward. PAN Communications found that overall media mentions dropped 41% year over year while brand reach increased 10%, because these systems prioritize context and credibility over quantity. A hundred pieces of generic content won’t outperform a handful of well-sourced, well-placed stories in outlets that matter. Flooding the zone with machine output may actually work against you as models continue to improve their filters.

4. Old Coverage Still Works for You

A story from two years ago in a credible trade publication can still shape what a model says about your company today. AI systems reward consistency. Repeated signals from multiple sources over time carry more weight than a surge of new content from one place.

Muck Rack’s research on AI brand visibility shows that most AI visibility for a brand comes from roughly 20 outlets, and that coverage patterns matter more than single placements. Companies that kept up a consistent PR program over the last few years have a structural advantage right now, and that gap takes longer to close than most people realize.

5. Your Owned Content Performs Better When Earned Content Backs It Up

A well-organized FAQ, a clearly written case study, a founder’s perspective with enough specificity to be quotable: these are worth producing. Models do ingest owned content, and structure and clarity improve the odds your material gets cited correctly.

The catch is that owned content performs better when third-party coverage backs it up. Fast Company captures it well: earned media is king, owned content is queen. A media placement links back to your site, signals to AI systems that your content is worth referencing, and gives the model the triangulation it’s looking for. The two work together, and the strategy breaks down when you pull one out.

A concrete example: a fintech company that earns a quoted mention in a Forbes piece on embedded finance, then publishes a follow-up post on their own site with more depth on the same topic, is more likely to show up in a relevant AI answer than either piece would be on its own.

The Starting Point

Before rebuilding your content strategy around what AI systems value, it helps to know how they currently describe you. Which outlets are they pulling from when someone asks about your category? Who in your peer group is showing up, and why? What narrative is already attached to your company name?

Most companies are surprised by what they find when they actually look. That’s where doing a proper audit comes in.

So What Should You Actually Do?

The practice of optimizing for how AI systems find and cite you has a few names floating around right now. GEO (generative engine optimization) is the one gaining the most traction. Whatever you call it, the playbook is more familiar than it sounds. Here are five things worth doing now.

1. Get a baseline. Before you can improve how AI describes your company, you need to know what it’s currently saying. Run prompts about your company, your category, and your competitors across ChatGPT, Perplexity, and Google’s AI overviews. Note what sources show up, what narrative gets repeated, and where you’re missing or misrepresented.

2. Map the outlets that matter for your category. AI models don’t pull from everywhere equally. Muck Rack’s research shows that most AI visibility for a brand comes from roughly 20 outlets. For your industry, there are probably 10 to 15 publications that carry real weight with these systems. Know which ones they are and build your media strategy around them.

3. Prioritize quotes and direct mentions over general coverage. A passing mention in a roundup is less useful than a direct quote attributed to your CEO in an industry analysis piece. Fast Company notes that AI models prioritize diverse, multi-source citations, and a concise, quotable statement from an expert in a reputable publication carries significant weight. Pitch with that in mind.

4. Structure your owned content so AI can actually use it. Clear headings, plain language, specific data points, and a consistent company description across your site, press room, and executive bios all help. LLMs extract passages, so every section of your website should be able to stand on its own as an answer to a question someone might ask. This is also where AI-generated content creates problems. Thin, generic, or repetitive material gets filtered out, and as models get better at detecting it, the shelf life of low-effort content keeps getting shorter.

5. Keep it consistent over time. A burst of coverage followed by months of silence is less useful than a steady drumbeat of well-placed stories. AI systems reward patterns. A company that shows up repeatedly across credible sources on the same topics builds a durable signal that’s hard for competitors to displace quickly. That’s the compounding value of a consistent PR program, and it’s more relevant now than it’s ever been.