As AI becomes the new gateway for how people discover and trust information, one big question remains: what content are these systems actually pulling from—and why? In this post, Muck Rack CEO Gregory Galant breaks down the findings from a new study that uncovers how AI chooses its sources, and what that means for PR teams and brand strategy. 

Over the past few months, the team at Muck Rack moved fast. Really fast.

As AI-generated content started to reshape how people find and trust information, we saw a huge gap: no one had clear, concrete data on what these tools are actually reading or why certain brands get cited in responses while others don’t.

So our team set out to answer it. And now we’re sharing the results. We just released a new study: “What is AI Reading?”

We analyzed millions of AI citations and tested thousands of prompts across ChatGPT, Gemini, Claude and more to understand how these systems source their answers and what it means for PR teams, brand visibility and strategy.

What we found confirms what many have suspected, but couldn’t prove:

Earned media materially influences what AI says. 

This isn’t just a matter of whether your brand appears in results. It’s about how it’s represented, what coverage it draws from and what AI believes is credible.

We saw that most citations (95%) come from unpaid sources, and over half of responses include at least one earned media link. And that AI's choices are shaped by factors PR teams can influence: recency, query framing and outlet authority.

If AI is the new front door for discovery, comms teams need to think not just about human readers, but algorithmic ones too. This study gives you a framework for doing that.

So we’re taking it a step further with Generative Pulse by Muck Rack, a new integration that helps PR teams uncover the sources AI trusts, earn their citations and stay in the conversation. Learn more about what we’ve been working on here.