76% of Domains Cited by AI Appeared Only Once

Lightly edited for clarity. Speaker labels retained.

Gini Dietrich: (cold open) PR teams are using AI to write more pitches than ever, and journalists are blocking them. While that's all happening, the AI engines people actually use to make decisions are citing sources most of us have never put on a media list. So, are we pitching the wrong places? That is what we're going to dig into today. I'm Gini Dietrich, I'm the founder of Spin Sucks and creator of the PESO model, and I'm here with Brett Farmiloe, the CEO of Featured, which is the company behind Help a Reporter Out, you all know it. Brett's team just classified about 23,000 AI citations, and some of what they found should make every single comms team rethink its media list. Brett, welcome.

Brett Farmiloe: Yeah. Thanks for having me.

Gini Dietrich: Of course, I'm very excited to have this conversation with you because it's a little mind-blowing, I'm not gonna lie. So let's start with your side of the house. More than one in five HARO pitches are now written entirely by AI. What does that look like from where you sit, and what are journalists doing about it?

Brett Farmiloe: It looks like a whole lot of automations. For people who are trying to increase their AI visibility, but might be doing it in an approach that values quantity and not quality. And it's a deep sigh on our end, because it causes a lot of work to make sure that the connection between journalists and sources is authentic and helpful. And so, when we first acquired and revived HARO, the very first decision that we had was, hey, we need to prioritize quality and trust in the system to make the revival a success. How are we going to do that? And we said, you know what? Why did it decline in the first place? It was a bunch of AI slop.

Gini Dietrich: Was it?

Brett Farmiloe: It was a very simple system because it's been around since 2008. It's email based. And if it's a simple system, we need to have complex guardrails to make sure that that quality and trust is gonna be at its core. And we said AI detection is one of those core elements. So we decided to partner with a company called Pangram and run every single pitch through Pangram to see if it was AI-detected. And we've scored, since revival, across HARO and Connectively, which we also revived, more than 1.4 million pitches, and we found that 21% of all of those pitches were 100% AI generated.

Gini Dietrich: I feel like 21% is low. Just based on the pitches I get? I feel like that's low.

Brett Farmiloe: Yeah, and those are 100% AI-generated. So we also see that on the other side of it, with journalists, very early, we heard feedback to say, hey, spare our inbox. We don't want AI generated slop. And so we include a little opt out of AI generated pitches on all HARO requests. We see about 35% of journalists click that little box that says, spare my inbox.

Gini Dietrich: Wow. That's not clicked by default, and really requires you to consciously press it.

Brett Farmiloe: And so you kind of see two different camps of journalists. You see the ones who definitely don't want any AI generated stuff. And then surprisingly, and probably more surprisingly on my end, it's the journalists who don't care if it's AI generated as long as it gets them connected to the right source for the story, in which case then they pick up.

Gini Dietrich: Interesting.

Brett Farmiloe: And then they do the interview, and then they do their job.

Gini Dietrich: So, that's interesting, because if you think about it, it's still requiring the human, right? So if we think about how all of this technology can make us more efficient and more productive and all of those things, what is the AI getting wrong that the human pitching gets right? And is there a solution in there?

Brett Farmiloe: We're finding that human-written pitches are 3 times more successful than AI-written pitches. And no surprise. The reason for that is because AI is the baseline, and journalists want more than the baseline. So if you're at the baseline, you're probably not gonna get over the finish line from a journalist perspective. So I think what's missing is that human written pitches are able to slow the pitch down, find out what's gonna win, and put that winning aspect into their pitch. Whereas AI is going to generate something that is going to be sounding great, and will meet the baseline, and meet the checks, and follow the instructions that the journalists wrote, but from where we're at today, at the end of 2026, there's still just something missing that doesn't get it over the finish line.

Gini Dietrich: I have, like, 17 questions in my head, hang on, I gotta filter. What I'm really curious about is, let's say that AI writes your baseline, which I think is probably okay, and you add on the extra stuff. Is that going to show up in your AI detection, because AI wrote the baseline?

Brett Farmiloe: It kind of depends on the input. And what I mean by that is we'll have people on a treadmill in the morning, getting their morning HARO edition, and they see it, and they voice their thoughts into AI, and then they have AI polish it and put it into that initial starting draft. That, from what I've heard and have discussed with our AI detection partner, will be marked as mixed AI generation. It won't be 100 percent AI generated. Versus if you have a prompt, and say it's a five word prompt, here's a HARO thing, reply to it. That'll probably be marked as 100% AI generated, even if you start to massage some things around and edit a word here. Like, you might be able to get by the AI detector, but what's really important is the actual input that goes into it to determine the output.

Gini Dietrich: It's like we could talk about with content. I mean, if you add your experience and expertise, you're going to win every time. Even if you use AI to draft the first piece, if you add your expertise and your experience, you're going to be differentiated. So it's kind of the same thing.

Brett Farmiloe: Yeah, exactly. And I think that that's what AI misses, is the experience. The expertise stuff, that's getting better and better, hour by hour. But the experience aspect of it, like, at the end of the day, it's like, what do readers want? They want something interesting to read and to consume, and the human behind it is the interesting component that makes the story different.

Gini Dietrich: Yeah, for now.

Brett Farmiloe: I mean, forever.

Gini Dietrich: I don't know, I just heard on the radio the other day, it's Friday, maybe, that the AI songwriters are getting so good you almost can't tell. Which is scary.

Brett Farmiloe: Yeah, I think it's different types of content though. It is a weird place that we're in.

Gini Dietrich: It's so weird, right?

Brett Farmiloe: Yeah, super weird. It is a weird spot, because, like, to your point, I was at, this is off topic or whatever, but it still illustrates it. I was at my stepbrother's rehearsal dinner. He's getting married.

Gini Dietrich: Very sweet, yeah.

Brett Farmiloe: And someone played a song. He kinda lied and embellished it a little bit. He's like, hey, you know, I got a friend in the music business. He wrote a song, and he played the song. And the wife was in tears. Oh my gosh, this is such a beautiful song. And I run into the guy. I'm like, dude, what a great gift, man. Like, that is so cool that you had your songwriter buddy write this song. It was AI. It was AI, and he admitted it to me, and the wife didn't know, the groom didn't know, everyone was caught up in the moment. But yeah, it's that type of world.

Gini Dietrich: It's insane. I mean, it's exciting, but also it's insane.

Brett Farmiloe: Insane.

Gini Dietrich: Okay, so walk us through the study. What did you measure, and what's the one finding that surprised you the most?

Brett Farmiloe: Yeah, so I think that you're talking about the AI citation study that we did. We launched a GEO audit product within Featured, and Featured is the copilot for PR. So it stacks all these different jobs to be done. GEO is one of 'em. And so we saw more than 400 GEO audits conducted. We analyzed a little over 22,000 citations. And the thing that surprised me the most was how long tail these citations are. We looked at it and basically said that, I think it was about 34% of all citations came from websites with domain authorities of under 40.

Gini Dietrich: Okay.

Brett Farmiloe: Domain authority is a score from Moz, essentially 1 to 100, how authoritative is your website? And so, these 34% of sites under a domain authority of 40 would never make it to a media list.

Gini Dietrich: Never.

Brett Farmiloe: This would never be, oh, we really need to reach out to this mom and pop blog and try to get them to cover things. And the reality is that the long tail of AI citations is so long, and it dramatically redefines what your media list actually looks like. So that was one of the biggest surprises that we saw, was the fact that the sites with authority and the ability to manufacture AI visibility are websites you probably have never heard of.

Gini Dietrich: So when you're thinking about your media list, then you really should… I mean, I keep saying, stop going after logos and prestige. Find the ones that are really valuable to your audience. You should be looking at, for lack of a better term, the Tier 3s.

Brett Farmiloe: I think it starts with an audit. It starts with what is AI actually citing about your space, your industry? And then it's like, okay, now that we have the sites that are being cited, what content is being cited? What are the content types? What are the places that AI is valuing in this? And then from there, it's figuring out your strategy. How are you going to get into that space? Are you going to create similar content? Are you going to pitch the sites that are already being cited and try to insert yourself into that? Are you going to try to spin it to collaborate on something new and different that's adjacent to it?

Gini Dietrich: Well, you're not, because spin sucks. But go on.

Brett Farmiloe: But all that to say, I think that it's a really exciting time that redefines what that pitch list actually looks like.

Gini Dietrich: So, okay, so talk more about that. If we're thinking about our media list, give me an example of when you would want to do a 39 domain authority versus a 90.

Brett Farmiloe: So, a good example would be, to dogfood our own products, we have a few different clients, so we kind of operate as an agency. One of those clients is Cruise America, and so they do RV rentals, and the pitch list is like these crazy weird RV type of blogs, like RV Pro. And RV Pro has a domain authority of 29 or something like that, but is an excellent source that gets cited all the time. And so you basically take a client like…

Gini Dietrich: People read it too, or is it just the bots that are reading it, the LLMs?

Brett Farmiloe: I would say I don't have access to their analytics, but my hunch would be the widely known stat that there's more bots reading internet content than people these days. I think Cloudflare cited that. So RV Pro, I would say, is publishing content majority for bot-related audiences.

Gini Dietrich: Insane.

Brett Farmiloe: And that's not to put RV Pro on blast or anything like that.

Gini Dietrich: No, sure.

Brett Farmiloe: But I think that that's probably the reality of today, is that publishers are producing for machines and people, and these long-tail publications are probably getting their leg up by focusing on having LLM-related type of well-structured content. And that puts a brand like Cruise America now on RV Pro's map, opening up an entire new business for publishers.

Gini Dietrich: So then, let's extrapolate that for anybody watching this. You extrapolate that to not only are you showing up in AI answers, but RV Pro is helping a client get to new audiences because they're recommending, oh, you should check out Cruise America, is that what it's called?

Brett Farmiloe: Yeah.

Gini Dietrich: Yes, like, you should check them out, and then it does all that work. It builds the shortlist that the humans aren't having to do anymore.

Brett Farmiloe: Yeah, and so that's really where the world has kind of shifted.

Gini Dietrich: What a world.

Brett Farmiloe: Yeah, you've got to do the audit to figure out who the heck is citing stuff in my space and industry, and then it's figuring out, okay, well, how do we work with these people?

Gini Dietrich: What a world. Okay, in the study, you said that 75-ish percent of domains were cited only once. So, when you think about that, are we having to build a gigantic media list and going after the domains that are cited only once, or are we trying to find where it's multiples?

Brett Farmiloe: I think this is actually probably good news for PR professionals, because that list gets longer and bigger as sites get cited. So I think that it just opens up the window of opportunity, where the top 100 domains accounted for about 25% of citations, so you're thinking about YouTube and Reddit and some of the usual players, whereas 76% of cited domains across our study appeared exactly once. And so what does that mean? I think it means that you've got to have the right tools to focus your efforts. You've got to be able to identify, okay, if this is going to be a one and done type of opportunity for us.

Gini Dietrich: Right.

Brett Farmiloe: And then you have to manufacture or redesign your standard operating procedures around those one and dones, knowing that you're going to have a very, very long tail. How do you make sure that, here it is on Monday, we're focused on this aspect for this client. Okay, Tuesday, we're gonna move into batch B. Wednesday, batch C. And actually have to work through that.

Gini Dietrich: It's interesting, because you're not gonna get the New York Times all the time, right? So, essentially, that could be a one and done. The Wall Street Journal might be a one and done, right? You may not get that more than one time from one organization, unless, of course, you're the White House or something like that, right? But exceptions to every rule, but same kind of idea. Everybody wants the New York Times, but you're not gonna be in the New York Times multiple times throughout the year.

Brett Farmiloe: Yeah, I think that the rewards are less from the one big hit and more from the breadth of different sites that you could actually be cited on.

Gini Dietrich: Okay, so if you're a comms leader and you've spent your entire career building a Tier 1 list, do you just throw it out?

Brett Farmiloe: No. I think on the flip side of this, so 34% of the sites that were cited have a domain authority of under 40. I think it was almost a mirror image for the sites that were being cited with a domain authority above 80. So your tier ones are also highly important in this. And that's not a surprise, because obviously you need… it's the same paradigm where Google needed to trust…

Gini Dietrich: Yeah. Yes.

Brett Farmiloe: AI is the same way. You've got to be able to trust the sites that have the authority and built-in history to it. So I think it was around like 33% or so were the tier ones that were being cited. And you follow the money, a lot of deals have been struck in terms of AI citations. And so this is actually another layer to it, but you might see certain sites cited in one engine like ChatGPT, but not in Perplexity. And you might see another site be cited really widely in Claude.

Gini Dietrich: Okay.

Brett Farmiloe: But not in another one. And oftentimes it goes down to business deals and licensing and things like that. It goes back to the audit, that if you're crushing it in ChatGPT, but don't have any visibility in Claude, how do you now focus on the sites that are being cited by Claude? So you get into the onion and there's a lot of opportunity.

Gini Dietrich: Yeah, and Andy Crestodina did his annual blogger survey, and for the first time in, I think it's been 13 years, something like that, he showed that blogging itself is not doing the work that it used to do, which I don't think is any surprise. But it also showed that you have this big disparity between the high domain and the low domain, and there's that middle layer, which I think is what you found, too. What do we do about the middle layer?

Brett Farmiloe: Well, I think the one thing that is interesting about all of this is that you gotta measure it. See what works, and then double down on the stuff that does. And do it until it doesn't work anymore. And I think that's a formula that gets us through the years ahead, because what worked yesterday might not work for tomorrow. I think that we had a moment in 2026 where we were all celebrating Reddit, and then all of a sudden Reddit citations dropped 80% overnight.

Gini Dietrich: It's because they got in a fight with ChatGPT. OpenAI, I think they got in a fight with. Which, again, is the licensing deals.

Brett Farmiloe: Yeah, right. So it might have been temporary, and might have been, oh, watch what we could do if you don't pay us, type thing. But I think that, in general, any good marketer is saying, okay, let's try the middle layer. Let's go make all of these different pitches. Okay, what's our success rate? Okay, now, off of that success rate, did these placements actually work? Oh, they did, and we could basically measure it by looking at the citation scores within our tools and modeling that we have. So, it's about testing, doubling down on the stuff that works, and being able to do that through measurement.

Gini Dietrich: How are you measuring AI visibility today?

Brett Farmiloe: Frequent audits. So, I think that there's a couple tools out there that are widely adopted, sites like Profound and Scrunch and a few others. So those are incredibly helpful for a baseline, as we try to make sense of it all. We use just Featured, to be honest. We just run GEO audits frequently on our own product. But increasingly, I think that it's just finding the tool that works for you, having some sort of baseline and some sort of consistent measurement, and then being able to adapt based off of the analysis that you're getting back.

Gini Dietrich: Yeah, I think one of the biggest challenges we've found is that, well, it's kind of the wild, wild west with tools right now, which is not a surprise, right? But the tools themselves are sort of making up the GEO, because they can't prompt the AI tool itself, and they don't know what the humans are actually asking. So their AI is making up prompts that may or may not be what humans are actually asking, and so there's this little bit of a gap. So I always say, if you have the tool, for sure, but also do a manual audit every couple of weeks, just to spot check, because I don't think it's totally accurate yet. And it shouldn't be, because it's all still brand new.

Brett Farmiloe: It's not only that, but it's also the technical aspect of it, of how AI engines differ from search engines. Whereas search engines, you pretty much had a whole ranking on things. And there might have been some variants where if you Google search from New York, you could be position 3.4, and if you did it from Scottsdale it could have been number 1, and all those different things. So there's not that much variance with the old way search works. There's a ton of variance with engines and AI. You could have 500 different results based off of the model that was called and the engine that was done and the time of day and all this different stuff. So that's why the tools are having a difficult time with it, because there's way too much variance.

Gini Dietrich: Yeah, it's really hard. I mean, even clients are like, okay, well, how do we measure this? And I'm like, I gotta spot check it manually, sorry, friends. We can use the tools, but we gotta spot check it, too.

Brett Farmiloe: I feel like the best measurement really is just how did you hear about us, in demos and purchases.

Gini Dietrich: Yeah. The qualitative, one by one interview.

Brett Farmiloe: Yeah. Oh, we found you on ChatGPT and we were searching for this and blah blah blah. So that, to me, is the biggest signal that we've got that's reliable.

Gini Dietrich: That's crazy. Now you just need the sales team to ask it.

Brett Farmiloe: Yeah.

Gini Dietrich: Which is the age-old issue. Okay, so if a team could change one thing this quarter, now that we're starting at the beginning of Q4, based on this data, what should it be?

Brett Farmiloe: I think that, this might not be a surprise, but it's that AI visibility is the North Star. I think that it's the way the world is moving, and really engineering your systems towards AI visibility, I think, is the consensus. I think for maybe the last six quarters or so, we've known that this wave is coming. And this wave has washed upon shore and it's here to stay. And I think that it's now re-engineering everything, from the way your team is built, to what your team does from an SOP perspective, to the day-to-day tooling that you've got. I think it's just a complete restructure in terms of AI visibility. AI visibility matters, and it will continue to matter for the next decade.

Gini Dietrich: 100%, I totally agree with that. So, AI visibility is the North Star. Do an audit to see who's already being cited, and if that would be valuable for your organization. And don't let AI write your pitches.

Brett Farmiloe: Yeah, exactly. And get quoted in articles. I think that that's the self-serving thing, but we've seen, because we're so close to it, we've seen some really cool research papers and academic studies that show, if you're quoted in an article… I think that there was one study that was done, it was, like, 41% more likely to be cited by AI, with an article that has a quote from an expert. That's something that we're seeing publishers do. These platforms that we have are getting widely adopted and more usage, not only on the source side of things that need AI visibility, but on the publisher side of things that are like, oh, we need to do a content refresh. Let's go find an expert to just include a quote in the story. So I think that it was pretty lucky to have gotten HARO back, and we're seeing that it's gotten value from both ends.

Gini Dietrich: That's fantastic. I mean, everybody's a big fan, so I get it. Okay, I have one quick lightning round question. I'll answer it, too. What is the biggest mistake PR teams will make with AI visibility in the next year?

Brett Farmiloe: Can you answer first? Or do I?

Gini Dietrich: I can answer first. I actually didn't think about it either. I would say not measuring it. I think you're right, there has to be a baseline, and we have to understand, we have to spot check it, we have to understand what's being said, and what's missing, and where that gap is, and how we can fill it.

Brett Farmiloe: I think that's great. It's a better answer than I was gonna say, which was that I don't know if you can make a mistake at this point. Just because it is so new.

Gini Dietrich: Yeah. You can make a mistake if you don't measure it and you don't know if your efforts are actually working. Or there's still lots of people that are like, oh no, AI is gonna take my job, so I'm just gonna bury my head in the sand. Let's not do that either.

Brett Farmiloe: I like that one too. See, we should just flip the mic here, 'cause those are good. Most people really do want to… in a lot of instances, they're letting fear overtake the action. And I think that there should be more excitement on this, even though it is very scary to not know what the future could hold. But embracing the excitement allows you to redefine what that future actually looks like.

Gini Dietrich: Absolutely, and it is exciting. I mean, it's crazy, but it's also super fun. I enjoy my AI tools very much.

Brett Farmiloe: There's not too many opportunities that we have as professionals to really be a part of the redefinition of the way things are done. So that's why, if you're just entering into the profession, you're gonna look back two, three decades from now and remember this. It's almost like the internet got created.

Gini Dietrich: Yeah, right. You're exactly right. Well, where can people find the full report?

Brett Farmiloe: They could go to Featured.com and they could find it in our press room. And then if anyone isn't subscribed to Help a Reporter Out, you could go to helpareporter.com, enter in your email address, and get three emails a day from us with journalist requests.

Gini Dietrich: Yeah, I mean, I don't know why anybody would not be subscribed to it if they do media relations, so subscribe to it, for sure. If you want to figure out where your brand actually sits in AI discovery and how to move up, we are running a live visibility engineering workshop. We'll put the link in the description for you as well. Brett, thank you. I feel like you and I could have a conversation for hours about this, but this is definitely the kind of data that our industry needs, and I appreciate you sharing it.