End-to-end outreach workflow
Table of Contents
- Only ~1% of people click AI citations, but 75% of buyers pick whichever brand AI mentions first in a shortlist.
- SEO fundamentals aren’t dead, they’re just ~80% of the picture now.
- LLMs form answers by pulling consensus from reviews, publishers, and other brands — which is exactly what PR is built to influence.
- Proprietary data wins citations.
- Distribution still matters. Great data with no syndication, email, or paid push won’t reach LLMs or audiences.
- Newsjacking and real-time commentary matter more than any single citation or mention.
This is Kevin Indig’s second time on our podcast (his first was all about website topical authority), and this time I came out feeling better about digital PR than ever before.
Kevin’s been publishing incredibly important AI-related research each week on his Growth Memo newsletter, which I highly recommend subscribing to if you don’t already.
We are going to focus on two of his latest studies that I think have a ton of crossover with our digital PR and link-building audience: one on ghost citations (when you get mentioned by an AI tool but not cited) and the power of proprietary research on AI exposure.
Digital PR matters more than you think.
Below is a slightly edited transcript:
What are “ghost citations”?
Kevin Indig (03:00)
A ghost citation is when an AI gives an answer and cites a brand as a source, but doesn’t actually mention that brand in the answer text itself.
What inspired the study was a few reports showing that people barely click on citations. The most prominent one is from Pew Research, published in 2025.
They looked at around 900 U.S. adults and almost 70,000 interactions with AI — a pretty robust sample.
What most people focused on was that when Google shows an AI Overview, clicks on classic results drop by roughly half.
But they also found that only 1% of people click on citations in AI answers at all.
So on one hand, everyone’s excited — “we’re being cited by AI!” — but on the other hand, people barely click on those citations.
I found the same thing in my own research: people barely pay attention to citations.
Without moralizing about it, I wanted to understand: how often do brands actually get cited, and how often do they get mentioned in the AI answer itself?
Vince Nero (04:19)
Good timing, too — last week Google’s John Mueller clarified some of the new AI-related data showing up in Google Search Console. They’ve added a kind of AI visibility metric — I think they call it “generative search impressions” — which shows up in the left-hand sidebar. My understanding is that a citation now counts as an impression, specifically within AI Mode and AI Overviews. Is that right, Kevin? Not Gemini, correct?
Kevin Indig (05:02)
Yes, you’re on the money. Not Gemini, as far as I know.
Vince Nero (05:11)
That’s part of a bigger conversation about Google’s messaging — that while you might be getting fewer clicks, the clicks you do get are supposedly stronger.
All of this makes the citations-versus-mentions distinction really important right now.
Which do you think brands should be paying more attention to: citations or mentions?
Kevin Indig (05:44)
On the Google statements — I think this is a much bigger deal than most people realize.
We’ve gone through a structural shift we can’t go back from: people getting answers directly in search results or through a chatbot like ChatGPT is now the standard expectation. That changes the environment enough that clicks really aren’t the right metric anymore.
To answer your question directly: in my research, AI mentions are far more important than citations.
My business partner, Eric Van Buskirk, runs a lot of these user-focused studies with me.
In one, we recruited a panel of 50 average U.S. consumers and studied how they make purchase decisions using AI.
We found that 75% of the time, when someone gets a shortlist from AI — say they ask Gemini “what’s the best laptop?” and get three, five, or ten brand names back — they choose whichever brand appears first.
People are highly sensitive to AI mentions and barely read citations.
The one caveat: some companies, because of their business model, will struggle to ever show up as a mention.
Review sites like TripAdvisor, Trustpilot, or G2 get cited heavily — that’s their whole value as a platform — but they rarely appear as the actual answer.
Most AI models, especially Claude, won’t say “according to G2…” — they’ll cite G2 as a source but answer directly instead. So there’s a real limitation depending on your business model. But for most companies, AI mentions are where the value is.
There’s also a growing body of research — some of it mine, and a recent Similarweb study — showing that brand appearances in AI answers have a real downstream business impact, even though attribution is still severely limited. That makes it something of a bet: you don’t know the exact payoff, but if this channel is as big as I think it is, it’s worth betting on.
Vince Nero (08:24)
That does seem like a tough problem for brands. I want to tie this back to digital PR and brand awareness generally. Are you seeing any correlation with other metrics that help brands understand the uphill battle here? Say you’re starting a small jewelry brand — would you even tell them to worry about citations yet, or just focus on building the brand naturally first?
If you’re a small brand, where should you start?
Kevin Indig (09:14)
Good point — and yes, I’d agree with you. The first priority is always product-market fit: a strong customer base, getting your name out there.
Early on, it’s really about attention and general awareness.
Once you reach a certain level of maturity, you can build out a ladder of leading and lagging metrics.
That’s what I do with clients: we know there’s a lagging, hard-to-measure downstream effect, but until we can measure that directly, we track leading indicators.
The most basic leading indicator is bot crawls — are AI bots even retrieving your content?
That’s the entry ticket.
From there, you look at visibility metrics: Google search rank (still important), AI mentions, AI citations — ideally measured relative to competitors as “citation share” and “share of voice,” with share of voice being the more important of the two.
In the middle of the ladder are quality metrics: Are you appearing in the right context?
Are you mentioned the way you want to be?
What’s your AI sentiment?
These are all quantifiable. And at the bottom of the funnel is the actual business impact, which is hard to attribute directly — so you rely on self-reported attribution, asking customers after purchase how they heard about you, and building a picture from there of how many customers are coming through different AI platforms.
Whenever I implement self-reported attribution with clients, the numbers are often surprisingly high — sometimes 10% of new customers say they came through AI.
That’s usually much bigger than what direct measurement shows.
Part of the danger with focusing too much on citations is that people are still thinking in terms of referral traffic — but referral traffic from AI platforms is tiny, maybe 1–3% of total organic traffic.
Yes, it converts better on average, but the volume just isn’t there yet.
I think people are stuck in a “traffic prison,” trying to attribute clicks in a way that badly undersells how big this channel can actually be.
Vince Nero (11:55)
The self-reported attribution point really resonates — we don’t even do that ourselves at Buzzstream, as much as I’ve pushed for it.
I think people are hesitant because they don’t want to disrupt sign-up flows. But understanding where your customers actually come from seems like the name of the game.
You can do everything right in the SEO/GEO game, but if people aren’t finding you through AI, that’s a real problem for a lot of listeners, especially small businesses.
It raises a chicken-and-egg question, too — are people not finding us on AI because we’re not doing something else right, rather than AI itself being the wrong focus?
You mentioned Google search rank — how important is traditional ranking still?
Without getting too deep into the “is SEO dead” debate — how important is classic ranking for the average user today?
How important is classic ranking for the average user today?
Kevin Indig (13:28)
It’s still important. I think of SEO — meaning search rank specifically — as the skeleton of the body that is AI search.
Without it, the body can’t stand upright; it’s critical for survival.
But it’s not where the story ends — there’s also muscle, tissue, blood, organs, all the rest.
There’s a lot of debate on LinkedIn and X — maybe I should spend less time there — between the “it’s all SEO, just make good content” camp and the “SEO is dead, it’s all GEO/AEO now” camp.
Neither is right.
The uncomfortable truth, as usual, is somewhere in the middle.
All the major LLMs still rely heavily on search rank — they ground their answers in search results to reduce hallucination, and for that you need to rank well.
There are open questions about how much of that grounding comes from Google versus Bing versus a chatbot’s own web index — ChatGPT is building its own, and Claude relies heavily on Brave.
But if I had to draw a Venn diagram, I’d say classic SEO and AEO overlap by about 80%.
That remaining 20% is new, and it can make all the difference — this is a game of very tight margins. So I try to stay open-minded rather than saying either “nothing’s changed” or “everything’s changed.”
It’s always somewhere in between.
Vince Nero (15:22)
Specifically for digital PR and link building — that’s core to Buzzstream’s audience — what’s your take on how important that work is for powering citations and mentions these days?
How important are links and digital PR for AI?
Kevin Indig (15:49)
It’s gotten much more important — and I’m not just saying that because I’m on your podcast; I’ve written about it.
Before the AI wave really kicked off around November 2022 or early 2023, the mental model in SEO was roughly: content matters most, backlinks help, user experience matters some.
Then the Google antitrust trial revealed that user behavior and experience actually mattered more than most people assumed, and backlinks were more secondary — a huge revelation at the time.
Then AI hit, and that whole conversation faded. Now: yes, user experience and behavior still matter for search rank, which remains the backbone of AI search.
But LLMs lean very heavily on third parties to form their answers, because they’re probabilistic rather than deterministic — they’re trying to build consensus from many sources.
When forming an opinion about a brand or category, they’ll pull from review sites, publishers, affiliates, social platforms, and other brands.
That’s exactly why PR matters so much — it’s the primary way to influence those other sources.
Vince Nero (17:22)
That’s a good segue into the second report — your two-part Growth Memo series on proprietary research and its power for AI exposure.
When I talk to our most successful digital PR customers, the common thread is almost always proprietary research or original data — something that establishes thought leadership and that journalists (and, to a lesser extent these days, bloggers) actually want to write about.
What is the importance of proprietary research in AI exposure?
Kevin Indig (18:27)
The TL;DR: companies that publish primary, first-party data tend to get cited more.
First-party data means looking at your own product usage or customer data and publishing it as a narrative.
We found that industry benchmarks specifically carry an outsized advantage — especially when framed around buyer questions.
If Buzzstream published a “State of PR 2026” report answering questions buyers actually ask — what PR campaigns are effective, what does success look like — that format gets disproportionate benefits.
It combines unique data, which LLMs value highly, with direct answers to the kinds of questions people are actually prompting AI with.
That’s not just an LLM effect — it’s also a classic search-rank effect, tied to the idea of “information gain.”
You want to offer something better than the countless other pieces of content on a topic.
It’s very easy to fall into the trap of just summarizing what everyone else says — and that’s a reliable way to get hit by one of Google’s core algorithm updates, which increasingly look for genuine newness and information gain.
That said, you still need topical relevance — some overlap with what the best existing content covers, simply for relevance’s sake.
But from there, you need something genuinely new that other content doesn’t have, and primary research is one of the easiest ways to provide that.
One exercise I run with almost every client: take inventory of what research already exists internally, published or not — product usage data, customer behavior, market data — and check whether it’s actually showing up in your content.
If not, we look at ways to layer that data into existing content through refreshes, which usually has a strong effect.
Vince Nero (21:31)
You mentioned the “statistics roundup” post format — probably the most common (and sometimes overused) version of this, where teams gather a hundred or two hundred stats from around the web.
The smart teams — HubSpot is a good example — build these from their own proprietary trend reports instead.
It sounds like you’re saying the proprietary piece is the real key here.
Tell me more about methodology and how the findings are presented.
Kevin Indig (22:36)
It’s an interesting question whether LLMs specifically recognize methodology markers as a sign of legitimate research.
I don’t have a controlled A/B test proving it, so I can’t say for certain — but it’s likely, because LLMs are trained on so much actual research.
Part of why AI Overviews show up so often in the medical space, for instance, is simply that there’s so much medical research for the model to draw confidence from.
So the more your content resembles an actual study, the more likely it is to get cited — that’s my best read of the data, though not definitive proof.
Transparency matters for trust reasons too — as someone who’s published studies myself, I’ve found that if your methodology isn’t clear, people notice and push back.
Believability and trust are critical in this AI-driven environment, so I’d strongly advise being as transparent as possible about your data and approach.
There’s also a broader question of effort.
I’ve found that the amount of effort that goes into a piece is disproportionately rewarded — not just in AI search, but classic search too.
I think as an industry, partly because of how easy AI makes content creation, we’ve gotten a bit too sloppy.
I’m not against using AI for content at scale, but there’s something real to the idea of effort.
Google has taken a stronger stance recently on “commodity” versus “non-commodity” content — a lot of people have seen the slide Danny Sullivan presented at a recent conference on this.
Google’s core updates increasingly reward content built on genuine firsthand experience and expertise, and penalize content that’s just summarizing top search results.
Vince Nero (25:00)
Is proprietary data alone enough for a small brand — say that jewelry startup again — or does distribution matter just as much? If a tree falls in a forest and no one’s there to hear it…
Kevin Indig (25:38)
Distribution is absolutely critical. If you have a strong website that LLMs crawl regularly, you’ll get pulled into their training and retrieval more naturally — but distribution also matters simply for brand awareness.
Ramp, the fintech startup I used to advise, is a great example: they have rich first-party data on card spend and tell great stories with it — comparing, say, how much companies spend on OpenAI versus Anthropic.
That kind of storytelling shows up both in LLM answers and classic search, partly because they distribute it so aggressively.
Syndication is very effective.
I’d also push more brands to think seriously about owned channels like email — most e-commerce companies collect emails from transactions but only send coupons or basic promotions.
I’d argue it’s far more valuable to send something with real editorial value, so that when you do have something worth distributing, that list is warmed up and ready. HIMS, which I worked with for a couple of years, does this really well — using email to distribute genuinely valuable content, backed by paid distribution: search ads, YouTube, social.
On top of that: syndication, advertorials, sponsored content, and influencer marketing.
I’ve seen bigger brands pay well-known accounts to share reports — a big multiplier, both for AI visibility and for the business itself.
Vince Nero (27:47)
That distribution piece is where people can lose the plot if they focus too much on citations or AI visibility and forget the actual customer.
You’ve shown that information gain and proprietary research matter for AI citations specifically — but I’d argue it matters just as much for overall brand authority, even if it’s harder to tie to one clean metric.
I’ve seen agencies build a lot of success around an ongoing branded report — a recurring “state of the industry” piece — or expert commentary tied to breaking news, using proprietary data to newsjack a story.
That puts you into the conversation and raises brand awareness in a way that goes beyond any single AI mention or citation.
How important is newsjacking?
Kevin Indig (29:12)
I love the idea of newsjacking, of inserting yourself into the conversation. Okta is a great example — they built a full in-house newsroom with six full-time journalists doing nothing but real-time commentary and second-touch media.
Brilliant idea, because attention really is everything right now.
“Attention is all you need” is, of course, the title of the research paper behind the Transformer architecture that kicked off this whole AI wave — but it also applies to how brands operate today.
Being part of the conversation matters, however you get there. I’m not saying that’s how it should be — just that it’s how it currently looks.
Attention is genuinely hard to earn and hold right now, because everything is competing for it — Netflix, Slack, YouTube, whatever people are reading.
People bounce around constantly, second-screen usage is everywhere. There’s real value in owning attention.
Vince Nero (30:44)
That’s a strong argument for digital PR being about more than just numbers.
There’s a lot of nuance in figuring out what’s actually newsworthy about your own company — a lot of people struggle with that.
Early in my consulting career, I’d have to tell CEOs, gently, “nobody’s going to care about this the way you do” — you have to shape it into something that actually resonates with customers or a broader audience.
Have you run into common mistakes brands make when trying to earn that kind of attention?
Kevin Indig (31:55)
Two mistakes really bother me.
First: access.
Companies have legitimate legal and regulatory reasons to be careful with product data, but many gate access to it far too tightly — no clear owner, no easy pipeline, no cross-team collaboration.
That’s a guaranteed way to fall behind.
The number one opportunity for most product and marketing teams is learning to actually leverage the product, customer, and market data they already have.
Second: lame takes.
Too many companies start from what they have rather than what the moment needs.
The skill to build is observing what’s happening in the world — both the broader news cycle and your customers’ world — and asking where you can credibly position yourself in that conversation.
Only after that should you ask what data you have, or need to create, to back up that position.
If you start from what you already have, you’ll usually just add noise — not interesting to your audience, not tied to what’s happening, and honestly not something you’d read yourself.
Speed, access, and judgment about where you can credibly join a conversation — that’s the foundation.
Execution follows from there.
Vince Nero (33:47)
I always push people to think about timeliness and emotional resonance — those two things together seem to be what earns real attention. And with proprietary research specifically, I think the finding has to be genuinely unexpected — journalists won’t pick up the obvious stuff.
Are you seeing more companies hire journalists in-house?
Kevin Indig (34:29)
Yes, and I think it reflects a bigger trend.
I’ve made this prediction at a few recent conferences: we’re seeing a shift from public media to corporate media.
It’s a tough environment for journalists right now — publishers are in crisis — so more of them are leaving traditional outlets to join companies as in-house, corporate journalists.
I like that model a lot, even though I wish public journalism itself were healthier.
Journalists bring skills that translate really well to AI-era content — leading with the actual point, for one, which resonates well with how LLMs summarize.
And sourcing: journalists are trained to find primary sources, which most corporate content teams simply don’t do — how many SEOs actually go find and interview a subject-matter expert?
Very few, even though it’s a genuinely good idea.
Look at the scale of investment here — OpenAI reportedly spent something like $100 million or more sponsoring the TBPN podcast, which is probably overpriced but shows where the trend is heading.
One of the Collision co-founders is hosting his own podcast now.
When a co-CEO of a massive company is spending real time on something like that, it tells you how important owning this kind of corporate media presence has become.
Vince Nero (36:24)
Great place to wrap up, Kevin. Thanks so much for your time — if people have questions, drop them in the comments. And I highly recommend checking out the Growth Memo newsletter; I’ll link it in the show notes. This was awesome.
Kevin Indig (36:44)
It was truly awesome for me too. Thanks so much for having me on — always a pleasure, Vince.
Check out the BuzzStream Podcast
