Some of Treble’s Austin-based crew recently caught a talk by Muck Rack’s Natan Edelsberg as he shared some insight into the company’s latest “What Is AI Reading?” report. The report debuted last year, and we always look forward to the newest version, but having Natan talk through the slides live and answer questions along the way added a new level of detail.
Here are some of the highlights from the session, along with perspectives from some of the Treble team members who attended.
The 2% Overlap
This was probably the biggest surprise (which may be why Natan put it up front in the presentation). When Muck Rack compared their listing of the most pitched journalists with those most cited by AI engines for a particular brand, the overlap was only 2% on average. Later slides that broke down AI engines’ favorite outlets and journalists by industry (tech, travel, healthcare, etc.) didn’t include the big names like the New York Times and Wall Street Journal (more on that later). Natan suggested that adjusting pitching practices could have a major impact on how a company shows up in generative search results.
Earned Media Remains on Top
In what has become a consistent finding since the first report in 2025, about 99% of links cited by AI come from non-paid media. The biggest part of that is earned media at 84%, which includes journalism, academic research, government and NGO sources, encyclopedic sites like Wikipedia, social platforms, and third-party corporate content. That leaves 16% for owned and paid media.
We were happy to see that as part of that earned media total, traditional journalism accounted for roughly 27% of all cited links. Sydney Miner, one of our senior account executives, said, “Even with the pressures facing today's newsrooms, trust in authentic journalism remains. These AI systems still lean heavily on citing original news sources because that's where true reporting lives. This shows how important earned media is for breaking through the AI noise.”
Different Engines, Different Worlds
Each of the big LLMs operates in its own information universe. Reddit is Gemini’s top-cited domain, but it gets little or no mention by ChatGPT or Claude. They all cite Wikipedia pretty regularly, but that’s about it for overlap. When it comes to media outlets, none of them rank in the top five across all three providers. That’s a strong indication that the outlets that drive AI visibility depend heavily on which AI their audience uses.
The Axios Style
Axios was the outlier among media properties. With its short-form, structured style, Axios’ stories are well-built to attract AI engines. It also helps that much of Axios’ content is freely available (no paywall) and that its sites aren’t heavily reliant on advertising. Natan used “open” more than once to describe the most AI-friendly media sites.
“The most interesting takeaway from the overview is that each AI engine cites a different set of sources, yet Axios is one of the few that all of them consistently draw from,” said senior account executive Laura Corona. “For companies looking to improve their GEO/AEO, this points to a continued need to prioritize earned media and make sure press releases and owned content are optimized for AI engines.”
How Does Tech Stack Up?
We were particularly interested in how tech sites fared in AI citations. Overall, trade publications were more often cited than mainstream news, but the titles were interesting. TechRadar is ChatGPT's number one technology domain, ahead of Wikipedia, Tom's Guide, and YouTube. Claude’s top domains were Quora, Wikipedia and TechRadar, while Gemini preferred YouTube, Reddit, and Medium.
“The most interesting part for me is how important Wikipedia, Quora and Reddit are for LLMs and making sure that your company message and branding are accurate and up to date across those sites,” said account manager Katie Anne Hayes. "The other interesting find for tech-specific outlets is the importance of TechRadar!”
Among the tech journalists most cited by AI engines were Jowi Morales of Tom’s Hardware, Cale Hunt at Windows Central, Joann Muller of Axios, and Craig Hale at TechRadar.
We’re looking forward to the next edition of the report later this year, but it’s clear that even as buyers increasingly use AI tools for research and discovery, the basics of PR remain important. Securing earned media placements, keeping brand narratives consistent across all platforms, and targeting the trade publications that feed specific LLMs are essential to keeping your brand visible where it matters most.
A big thanks to the Austin chapter of PRSA for organizing the event, which leads us to our final wrap-up from media strategist Josh Georgiou:
As a first-time PRSA event attendee, seeing the broad spectrum of public relations professionals, from eager students to seasoned in-house and agency teams, was energizing.
I had a great conversation with an attendee named Stephanie, and our paths highlighted just how vast this field really is. Stephanie shared that PR had been her dream job since her student days, but after giving agency life a shot, she realized it wasn't for her. However, she discovered along the way that she thoroughly enjoyed writing, which led her down a different career path within the industry. My journey was almost the exact opposite: I knew I wanted a role in communications or marketing, but I didn't know anything about PR until my first job at an agency. Once there, I fell in love with Tech B2B PR and agency life and have stayed in it ever since. Seeing how the two of us ended up in the same industry but thriving in completely different environments was such a fun reminder of how many avenues PR offers. It was equally inspiring to watch college students eagerly dive into tools like Muck Rack and discuss AI’s impact on journalism. Five years into my PR career, their enthusiasm was a great reminder of just how cool this work is. To top it off, the Q&A provided actionable insights into how AI digests information, specifically how Axios’ easy-to-read format, complete with big-picture overviews and bullet points, makes it a frequent citation favorite for AI.