TL;DR

PR teams have traditionally built media lists around reach, reputation, and authority. But new research from Featured suggests AI engines are changing which publications matter. About 34% of the AI citations analyzed came from sites with a domain authority below 40, while 76% of cited domains appeared only once. Tier 1 media still matters, but so does the long tail. Communicators need to rethink media influence in terms of what both people and machines trust.

Key Insights

• Featured analyzed more than 22,000 AI citations and found that about 34% came from sites with a domain authority below 40.
• 76% of cited domains appeared only once, revealing a substantial long tail of AI citations.
• Human-written HARO pitches are three times more successful than AI-written pitches, while 21% of pitches analyzed were classified as fully AI-generated.
• Tier 1 publications remain important, but niche and trade publications deserve greater attention.
• AI visibility requires continuous measurement because different engines cite different sources, and those patterns change.

Are PR Teams Doing Earned Media Wrong in the AI Era?

AI has changed the communications industry in countless ways, particularly earned media. And it’s no longer a secret: 76% of respondents in a poll during our recent Visibility Engineering™ webinar said earned media drives AI visibility.

Data like this, along with many conversations about the topic, has me wondering: What if we’re pitching the wrong publications?

Gini Dietrich and Brett Farmiloe, CEO of Featured, the company behind Help A Reporter Out (HARO), recently discussed that very question. 

Brett’s team analyzed more than 22,000 AI citations across more than 400 generative engine optimization (GEO) audits. Some findings challenge how PR teams have built media lists for decades.

Their conversation began with AI-generated pitches, but what caught my attention was something bigger: AI may be changing what a valuable media placement actually looks like.

Here’s what stood out.

More Than One in Five HARO Pitches Are Entirely AI-Generated

When Featured acquired and revived HARO, one of Brett’s priorities was rebuilding trust in the platform. That meant confronting what he believes contributed to its decline: AI slop.

HARO is simple. Journalists request sources, experts respond, and connections happen. But maintaining that system now requires more sophisticated guardrails.

Featured partnered with AI detection company Pangram to analyze pitches submitted. Across more than 1.4 million pitches submitted through HARO and Connectively, 21% were classified as fully AI-generated.

Gini’s immediate reaction? That sounded low.

Mine too.

But Brett clarified that these weren’t pitches where someone used AI to improve a sentence or organize an idea. They were classified as entirely AI-generated.

Journalists have noticed. HARO now lets them exclude AI-generated pitches, and roughly 35% actively select that option. The remaining 65% don’t necessarily object to AI involvement. They want access to the right experts and useful information.

Which raises a question: If AI makes pitching faster, why aren’t AI-written pitches performing better?

Human Pitches Are Three Times More Successful

According to Brett, human-written HARO pitches are three times more successful than AI-written pitches. His explanation is straightforward: AI gets you to the baseline. Journalists want something beyond it.

An AI-generated response can follow instructions, answer questions, and sound polished. But everyone can do that now. What differentiates a pitch is the experience behind the answer.

A human can ask: What have I seen that someone else hasn’t? What can I tell this journalist that won’t appear in the next 20 pitches? 

Gini connected that distinction to content creation more broadly. AI can produce a first draft, but personal experience and original insight make it valuable. AI is increasingly capable of reproducing expertise. Firsthand experience is harder to replicate.

And that distinction becomes especially interesting when we consider which sources AI engines themselves choose to cite.

The AI Citation Long Tail

PR teams have spent decades organizing media lists around reach, authority, and prestige. A publication with a domain authority of 29 probably isn’t what executives ask about on Monday mornings.

But Featured’s research found that approximately 34% of AI citations came from websites with a domain authority below 40.

Think about that.

Many of those publications might never make a traditional media list. Yet they appear in the sources AI engines use to generate answers. As Brett explained, some websites creating AI visibility are ones communicators have probably never heard of.

That doesn’t mean abandoning established media lists or pitching every niche publication we find. It means understanding which sources AI already cites before deciding where to focus.

Start With What AI is Already Citing

Brett recommends beginning with an audit:

• Which sources are AI engines citing in your industry?
• What content are they using?
• Which publications consistently appear?
• Where are your competitors being mentioned?

Those answers can help determine where to pitch, what subjects to prioritize, and which publications deserve more attention.

Brett used Cruise America, a Featured client, as an example. The RV rental company’s citation landscape includes specialized publications such as RV Pro. With a domain authority of around 29, RV Pro might not attract much attention on a conventional media list. Yet Brett identified it as a frequently cited source in that category.

Gini asked a great question: “Do people read it, too, or are the bots reading it?”

Increasingly, I’m not sure that distinction matters as much as it once did.

A niche publication may have limited reach compared with national media. But if AI engines use their content to answer relevant customer questions, it can become another pathway for potential customers to discover a brand. That changes how we should think about media influence.

Seventy-Six Percent of Cited Domains Appeared Once

Featured found that the top 100 domains accounted for approximately 25% of citations, while 76% of cited domains appeared exactly once. That’s a remarkably long tail.

Gini wondered whether PR teams now need enormous media lists containing publications they might pitch only once. Brett’s answer focused less on list size and more on building systems to identify and pursue specific opportunities.

Instead of repeatedly targeting the same publications, communicators may need to find relevant sources across a much broader information ecosystem.

But there’s an important distinction.

A publication appearing once in an audit doesn’t automatically make it a worthwhile target. Relevance, credibility, audience, and the opportunity to contribute something meaningful still matter. And none of this means abandoning Tier 1 media.

Tier 1 Earned Media Still Matters a Lot

Highly authoritative publications continue to play a significant role in AI citations. Featured found that sites with a domain authority above 80 accounted for a substantial share, nearly mirroring the contribution from sites below 40.

That’s consistent with the broader idea that AI engines need signals of trust and authority. The emerging media strategy isn’t Tier 1 versus niche publications. It’s both.

What complicates matters is that AI engines don’t necessarily cite the same sources. A publication frequently cited by ChatGPT might appear less often in Perplexity or Claude. Licensing arrangements, retrieval methods, and other factors may contribute to those differences.

We already know earned media is important to AI visibility. The next challenge is understanding which earned media matters, to which engines, and for which questions. That requires a different approach to measurement.

Measure It. Then Keep Measuring It.

If one theme kept resurfacing throughout the conversation, it was measurement.

Brett’s approach was simple: Try something, measure whether it works, double down when it does, and adjust when it stops. It may be the least exciting advice in this article, but it’s probably the most important.

The AI citation landscape changes quickly. Brett pointed to Reddit as an example. Its prominence in AI citations fluctuated significantly during 2026, demonstrating how quickly assumptions about influential sources can become outdated.

Gini raised another challenge: AI visibility tools don’t necessarily know what actual customers type into ChatGPT or Claude. Many generate prompts to estimate what users might ask. That makes the results useful, but imperfect.

Her recommendation? Use the tools, but manually spot-check the findings every couple of weeks.

Unlike traditional search rankings, generative AI responses can vary substantially across models, locations, contexts, and even repeated queries. There isn’t a clean equivalent of the old Google rankings, and there may never be.

Which brings us back to a surprisingly simple measurement tactic: Ask customers how they heard about you. If someone discovered your company through ChatGPT, that’s valuable information. The challenge is getting sales teams to ask consistently and capture the answers. Some things in marketing never change.

For communicators, the goal shouldn’t be to chase individual citations. It should be to understand whether our efforts are improving visibility, credibility, and ultimately business outcomes. And that means connecting earned media to the rest of our communications system.

A placement shouldn’t be the finish line. It should feed owned content, shared distribution, and other opportunities that reinforce the same expertise. That’s where the PESO Model® becomes especially relevant.

The Biggest Mistake? Sitting This Out.

When Gini asked Brett what PR teams should change this quarter, his response was direct: “Make AI visibility the North Star.”

I wouldn’t interpret that as abandoning everything else to optimize for machines. But AI visibility can no longer remain an experiment sitting off to the side. It needs to influence earned media, content planning, measurement, and the broader communications strategy.

Gini’s advice: Audit the sources already being cited, determine which of those sources matter to your organization, and don’t let AI write your pitches for you.

Brett added one more: Get quoted.

Gini also asked what the biggest mistake would be over the next year.

As Brett asked for a minute to think about that one, she offered one herself: “Not measuring it.”

Without a baseline, organizations won’t know how they’re represented in AI answers, what’s missing, or whether their efforts are making a difference.

She raised another possibility: letting fear prevent teams from doing anything.

That one resonates with me. The technology is evolving, measurement remains imperfect, and the rules aren’t settled. But uncertainty isn’t a reason to wait. Brett compared this moment to the emergence of the internet. We don’t know exactly where AI discovery will lead, but we know the way people find information is changing.

So, are PR teams pitching the wrong publications? Not necessarily.

But we may be working from an incomplete definition of which publications matter. Tier 1 outlets still matter. Trade publications matter. And niche sites we once overlooked may matter more than we realized.

Our job has always been to understand where trust is built and how information moves. Now there’s another audience influencing that process: machines.

What to Do Next

Start with your media list. Compare it against the sources AI engines cite when answering questions about your company, competitors, and industry.

Are those publications on your list? Are you appearing in them? And when you earn coverage, are you connecting it to the rest of your communications program so visibility can compound?

We now have to build visibility for two interconnected audiences: the people we’re trying to reach and the machines increasingly influencing what they discover and trust.

The PESO Model Certification® shows you how to connect earned, shared, owned, and paid media into a system that builds authority and makes your expertise easier to discover.

And if you want to apply this specifically to AI visibility, join our hands-on Where Your Brand Sits in AI Discovery workshop on Thursday, October 22, from noon–2 p.m. ET. You’ll assess where your brand stands today and build an owned and earned plan to move from mentioned to cited to the answer. The workshop is $495, and seats are limited.

© 2026 Spin Sucks. All rights reserved. The PESO Model is a registered trademark of Spin Sucks.