End-to-end outreach workflow
Table of Contents
- Prompt volume doesn’t matter — what matters is whether AI understands your brand correctly.
- The buying decision happens before the click — AI does the research, so traditional funnel metrics matter less.
- Topical presence beats share of voice — being mentioned accurately matters more than being mentioned often.
- AI hallucinates about brands ~8% of the time — uncorrected errors compound over time.
- Third-party citations shape AI’s view of you — consistent messaging across PR, your site, and social matters more than ever.
What matters in AI and GEO these days is that LLMs describe your brand accurately in the context you need. Instead of asking “how do I rank for this keyword,” the smarter question is now “does AI actually understand who I am and what I do?”
In this episode, I sat down with Dixon Jones, longtime Majestic brand ambassador and now CEO of Inlinks and his new(ish) AI tool Waikay. Since his tool helps users understand what AI knows about your brand (Waikay stands for “What AI Knows About You”), I thought he’d be the perfect person to interview for this topic.
We talk a lot about his metric “topical presence”, but also things like brand hallucinations, why third-party citations and digital PR matter more than ever, and what it actually takes to “turn the ship” once AI has the wrong story about you. If you work in digital PR, link building, or brand marketing, this one reframes a lot of assumptions worth reexamining.
Below is a slightly edited transcript:
What does Waikay mean?
Dixon Jones: Let’s start with the name. I appreciate that people out there can’t pronounce it, can’t spell it, and we don’t have the dot-com.
There’s a lot that goes into it. But until you know that Waikay stands for “what AI knows about you,” it’s one of those brand names where you don’t know why it’s there.
As soon as you know that, the plan was that you’d talk about it on a podcast, in a lift, or at a conference and say, “It stands for what AI knows about you.” Job done. That was the plan, anyway.
You’re absolutely right that when we launched a year ago, lots of other tools were launching too.
I think Profound was already out, with a bunch of others jumping out maybe a little after us. But the vast majority decided to take rank checking and turn it into a tool for AI—”rank checking for AI” is basically what the developers were all thinking about. That wasn’t what we were thinking about.
What we thought about was this: somebody is going to come along and ask, “What’s the best vacuum cleaner brand?”—or whatever the question is. And they might not even type anything.
They might get an email that asks a bunch of questions, then press a button in Outlook and say, “Reply to this email with some ideas.”
So my thinking about how people use LLMs wasn’t based on how many people type a particular question in a week or a year—you’re never going to track that in a month of Sundays.
What we really needed to understand is whether the AI understands your business and brand correctly in the context of what you do, so you have a real possibility of being in the response in the first place.
So we didn’t start with a prompt-tracking system asking “What’s the best vacuum cleaner brand?” 500 times and seeing how often you appear.
What we did first was ask all the AIs, once: “What do you know about”—say Dyson, or Shark—”in the context of vacuum cleaners?” and let the AI fully spill the beans on what it knows. Beforehand, we’ll have gone to your website and seen what you say about your brand in the context of vacuum cleaners.
Then we create a knowledge graph of topics from the AI’s response and from the website, and match them up to score how accurately the AI describes each topic in the context of your brand.
We didn’t start from “more people ask about vacuum cleaners than about suction for carpet cleaning.” We started from the fact that your business is unique to you.
You’ve defined it on your website—you’ve potentially spent billions telling your story there. Let’s not build a tool that throws that out the window; let’s build one that works from what your business already is.
So all the recommendations extrapolate from where the AI has misinterpreted or misunderstood your brand.
It was only later, when we saw competitors doing prompt tracking—how many times your brand shows up in the context of “best vacuum cleaner” versus other brands—that we built a prompt analysis tool too.
But we kept the knowledge-graph idea. We analyze the output of the prompts: when you ask “What’s the best vacuum cleaner?” we turn the answers into underlying topics, so we can see which topics Dyson is associated with, which topics Shark is associated with, and so on.
That lets us take any prompt and create this idea of topical presence.
Share of voice is how many times your brand comes back compared to all the other brands—divide one by the other. Topical presence is how closely and how well the AI understands your response and content relative to the other brands mentioned.
So if Dyson always comes back in the context of suction, and Shark always comes back in the context of pet hair and de-tangling, you start to see the gaps.
The topic reports push people from the start toward better topical presence, and the prompt analysis shows where your topical presence is weak and what you need to cover more of to be better understood by the LLMs.
Will AI ever offer us something like search volume?
Dixon Jones: I think people like Profound are trying to do that already, using clickstream data or whatever their secret sauce is, to give you numbers.
But the numbers are entirely unrealistic, and it doesn’t take a genius to see there’s a flawed methodology—”we’ve got a data source, let’s go with it.” That’s not how people use AI anymore.
Let me change the analogy from vacuum cleaners to trainers.
Most businesses—not SEOs—sell things. They make things and sell things. Only news outlets are really selling information; the rest of us are selling a product, or a SaaS technology that helps make sense of the world’s information.
The reason search or click volume matters less is that the user is trying to get the AI to do all the legwork. There’s no click. The person asks the AI a question, the AI goes out and looks at your website, your competitors’ sites, review sites—whatever it is—synthesizes all that in seconds, and comes back with a coherent argument for where you should go or what you should buy.
There’s no click involved.
Think of that person as a VA, a virtual assistant, a real human being, or an AI agent—they’re doing the same thing: producing a coherent argument. It’s as if you’re the managing director and you’ve asked your PA to go find information, and it’s happened in a second. So click volume is completely pointless.
The buying decision is made before the click, because they’ve already seen all the arguments for why to buy your product versus someone else’s.
That’s why the way people consume AI matters so much. Take the example of someone just pressing “respond to this email.”
As far as the human is concerned, the prompt is “reply to this email”—there’s no obvious relevance. And a lot of the best intel from AI comes when you tell it to summarize information you’ve given it.
You upload a file and say “read this,” or “summarize this”—you’re adding your own data layer, and the AI is really, really good at that. Until we understand that this is what the user journey looks like, we don’t understand that search volume means nothing.
Vince Nero: Yeah, we’ve had internal conversations about this. Before I was even hired at BuzzStream, I remember the guys calling me—I was at Siege Media at the time—asking for my take.
The thinking then, and it’s kind of true now, is that people are just far better informed as buyers because they’re doing all the research, and then they arrive at your website much more informed. To your point, they’re doing all that research on AI beforehand, so the top of the funnel is almost out the door.
All that search volume doesn’t really matter anymore.
Dixon Jones: Right. Our philosophy has been: let’s make sure the AI understands us correctly.
I may be the smallest cog in the wheel, the smallest brand, but as long as the AI absolutely understands who I am and why I’m there, I’ve got as much chance as anybody of being surfaced in the right context.
When someone asks a more complicated question, the smaller and more focused the brand, the better—if they’re profitable, they’re focused on a narrow vertical.
A small brand isn’t selling every flight on every airline; it’ll get beaten by Expedia.com, for instance.
So a small brand has to be narrow in its story, and the better it explains its own DNA to the machines, the more likely it’ll surface. It’s not about the traffic anymore. It’s about the sale.
Vince Nero: Yeah, for sure.
It’s almost like you want to be there for the moment the person is ready to make a choice. Even if you’re very niche—say the question is “I want black trainers with solid arch support for lower back pain”—you want to be the one that surfaces.
So AI isn’t really a brand awareness vehicle; it’s more of a… I don’t even know.
Dixon Jones: Product differentiation tool.
We’ve got to focus as brands on why we’re different from the other guys, not why we’re the same.
Vince Nero: Love that.
I want to tease out another thing you mentioned, because it’s important for the digital PR and link building listeners: having a uniform brand message across all your platforms.
That feels like the crux of what Waikay provides—helping you understand where the differences are, where you position yourself as “the trainers for lower back pain” but the AI models see you as, say, the best trainers for teenagers—something completely different from what you’re putting out.
Dixon Jones: Yeah—or for running the New York Marathon.
Vince Nero: Right, something completely different.
That keeps coming back.
I was just at a round-table with the Reboot team in Manchester, and that’s where we ended up too: you have to be consistent across everything. Your LinkedIn profile needs to say what you want, and what people say on Reddit needs to mirror it.
So my question is: what are some solid, realistic steps you can take when you see—in the tool or on your own—that you want to be known for one thing but aren’t?
I run into this at BuzzStream. We want to be known as a digital PR tool, but we’ve existed for 15 years more as a link building tool, so we’re still on every listicle as a link building or outreach CRM.
If we want to reinforce that we’re a digital PR tool, what are some realistic things to do?
What can brands do to build topical authority?
Dixon Jones: I’m not pretending we’ve got it perfect yet, but right at launch we built a hallucination, or fact-checking, module—we did that instead of a sentiment analysis score. (We can talk about why we didn’t launch sentiment analysis if we have time.)
We ask the AI what it knows about a brand in the context of a product, let it fill out all that information, and then reduce those free-form answers into facts—one-liners.
“This one is prominent in the UK.” “This one does that.” That makes it very easy for a human to go down the list and check: is that right? Is that right?
From some internal research, I’ve heard hallucinations run around 8% on average—so there’s a lot of brand hallucination out there.
I’ll pick on Inlinks first.
We had one where the AI said Inlinks is a rank checking tool.
Inlinks is not a rank checking tool—it’s entity SEO technology; it doesn’t check rankings.
We traced it back to one positive review of Inlinks that had added a feature we don’t actually have.
We don’t want to claim we’re rank checking software if we’re not, because that dilutes the brand equity around what we do.
In this case we simply contacted them and pointed it out; they were delighted to improve their content, and we were delighted to find it—so they got a link out of it too.
Another good example—and I apologize to Jason Hennessey for this one, but I keep coming back to it.
When I first launched this, I was at a conference in Phoenix showing it to Jason, who runs a digital agency, hennessey.com.
Back in the SEO days, there was a company called something like Hennessy Auto Parts, and to differentiate he convinced them to put a link on their site saying, “If you’re looking for SEO, go to hennessey.com.”
Now, his is Hennessey with an “ey,” but there’s also Hennessy brandy, spelled without that “e.”
So when you ask, “What do you know about hennessey.com?” the AI came back saying, first, it’s an auto parts manufacturer—citing that site—which said if you want an SEO company, go there.
It got confused by the SEO reference on the auto parts site, and it was also confused with the brandy, which isn’t surprising because it fixes typos at the query phase, not the analysis phase.
So fact-checking is the answer to your question.
Hallucinations need to be fixed, because if you let a false narrative get out there in the LLMs and don’t address it, over time it keeps propagating a miscommunication—and it’s hard to turn the ship.
For BuzzStream, for Inlinks, for any of us in SEO, if the AI already has legacy training data saying you were something else, it takes effort to move.
The more we get it right in the first place, the less we have to turn the ship.
But we will have to move as brands, so we need to be able to see what the AI is saying as we do.
Vince Nero: One of the things I wanted to ask about—and the reason I reached out—was a post on the Waikay blog about your metric of AI topical presence.
It’s a way to understand how AI views you as a brand, so you can correct course if it’s veering into a different topic than you want.
Can you tell me more about topical presence?
Dixon Jones: Sure. Topical presence works around an individual prompt. Having built a product that ignored individual prompts while the rest of the world did prompt tracking—counting how many times your brand is mentioned regardless of volume—we needed something that competed with that.
So we do something similar to everyone else: you choose a cadence (a day, a couple of days, a week) and we ask the question across all the different AIs.
We automatically detect the brands in the responses—a lot of tools make you list all your competitors, but we find them automatically.
So take “What’s the best trainer for the New York Marathon?” First—is a “trainer” a person coaching you, or something you put on your feet? Big question.
But say it’s the shoe. It comes back with a bunch of brands, and if you ask that every day for a week across, say, ten LLMs, you’ve got 70 answers in one week.
If each answer mentions about five brands, that’s 350 brand mentions—some repeated.
Now you have a share of voice: how often your brand is cited compared to all cited brands, and it’s not confined to competitors you nominated.
That matters, because if you only ever list a couple of competitors, you can artificially inflate your share of voice. You don’t get to choose your own competitors anymore.
Topical presence is the other missing factor.
When you ask “What’s the best shoe for the New York Marathon?” some topics become more important than others.
Back pain might not matter for that occasion, but durability might—and concrete might suddenly become an important topic, because that’s the surface your feet are hitting.
When we see the topics coming back for that specific prompt, we compare them to the knowledge graph of topics on your website—your footprint of who you are.
So topical presence measures the breadth, depth, and consistency of the topics you talk about on your website versus the breadth, depth, and consistency of the topics the LLMs spit back about you.
It’s out of 100, and it applies to each individual prompt at a point in time—each prompt has its own topical presence score.
Vince Nero: That was going to be my follow-up.
Is this the kind of thing where you can find yourself in a hole—going deep on a specific prompt type and missing a bigger picture?
How do you recommend people build these prompts?
Dixon Jones: The way we do it: topical presence is by individual prompt, but if you’ve got 100 prompts we’ll bag them into groups.
You can tag them—these are the New York Marathon prompts, these are the back pain prompts, these are the price prompts.
It’ll auto-tag, but you can change the tags.
Then you can take a basket of questions people might ask and see what consistently comes back for each basket.
But this is the start of a journey.
Having a metric for the accuracy of your topics is really just the start, because it shows you where you’re missing coverage—coverage in the LLM.
If the LLM doesn’t know your shoe is good for a hard-wearing surface, maybe you haven’t talked about it on your website, or maybe you have and it didn’t pick it up. Either way, that’s the hole.
So you make sure your website covers it. Ultimately it may be that a third-party website is where it needs to be—and that’s where you’d use BuzzStream tools to get your message out to third parties.
That’s the magic bit Waikay doesn’t do. But it’s what’s in the message that matters: if you need to talk about concrete, talk about concrete.
You wouldn’t necessarily have known that until you saw the LLM saying the New York Marathon is all roads, so you need good road-wearing shoes. “What’s a road-wearing shoe? Let’s go find out.”
Vince Nero: That’s a fantastic segue into digital PR and link building.
You touched on it, but can you elaborate on the role of third-party coverage—specifically digital PR—in increasing your messaging and getting your brand associated with the right keywords and messages?
Why is third-party coverage – specifically digital PR – so important?
Dixon Jones: It’s really interesting right now. As we’re recording this—it’s April 26th, 2026—over the last twelve months we’ve noticed a huge re-ignition of listicle sites: “Here are the top 10 Christmas card companies,” whatever it is.
Those listicles act as a third-party reference for the LLMs.
If the LLM understands your business properly, that’s great, but what it likes to do, when it cites anything, is cite a third-party source rather than your own website—which is a bit frustrating for the brand.
It’s not black and white, but if your brand is cited in the right place alongside your competitors out in “news land,” it definitely increases your chance of being cited in the response, because the LLM likes third-party material.
What went wrong for the LLMs is that they don’t have anything like PageRank to hand—maybe Gemini does.
I think they’ve over-relied on the assumption that all this content is equally valuable.
A listicle I wrote about Christmas cards because I had a Christmas card client ten years ago, versus an article I wrote about SEO where I’m known and have some authority—those have totally different values.
One’s authoritative; one’s Dixon trying to spam the internet.
They didn’t have a real weighting metric.
I keep hoping someone from the LLM world says, “Let’s get all the data from Majestic,” because Majestic has this brilliant proxy for PageRank—but we’ll see if that ever happens.
I’m not part of Majestic anymore, so I don’t have any secrets; I’m just a Majestic ambassador.
Either way, third-party citations really help your chances of getting into the results.
But ultimately, the context of your own story is how you control that narrative.
Vince Nero: Right. To bring it home and back to what we talked about at the beginning: if you’re pushing out messaging, even in the news—a lot of digital PR strategy relies on things like city index studies, “the best cities for X,” compiled by some website—the piece people sometimes miss is the contextual messaging around your brand in any campaign.
The way I’ve understood it: if I’m getting a link or a mention, I’d want to push the journalist or blogger to say “digital PR tool BuzzStream” every single time—and not just there, but to match it on my own properties, always mentioning those things together so I own that association.
That’s topical presence within that topic: if people are talking about digital PR tools, that’s what you want to be known for.
Dixon Jones: I’d see it like this: you need the citation to get the hook—for the LLM to put your brand in the list.
Then, as a human, they’ll say, “Tell me more about Waikay,” or “Tell me more about BuzzStream and this digital PR element.”
The user digs.
Once you’re in the list, it becomes a case of “Okay, I’ll check all of these and see which ones do digital PR, which do link building, which do road-running trainers.”
And once it’s into your ecosystem—your website—and understands you, if your story is clear, you’ll do all the things your salesperson was supposed to be doing for your brand.
It becomes a place of clarity.
But you’ve got to get through that initial leap of faith, which is hard for smaller brands, because the AI did its training data a long time ago and uses RAG—retrieval at the last minute—to pull in new sites.
The hook is usually in that RAG element, the search lookup.
But once we’re there, the AI dives into the website and pulls all the salient information. That’s where we win.
Vince Nero: This has been great. I think we should close it there.
I want everyone to go check out Waikay if they haven’t, and check out Inlinks too.
If we’re talking about AI, getting everyone at your company on board with putting forward the same messaging across all aspects seems more important than ever.
If you and I were just talking about marketing or building a business in general, it’d be obvious—but nowadays, especially at a big company, you’ve got one team pulling this way and another pulling that way, and it’s really easy for the messaging to get pulled apart.
A tool like Waikay can highlight that and help you get buy-in for the difficult conversations agencies or in-house teams are having, by pointing to the gaps and saying, “Hey, we need to work together, because we’re all saying different things.” To your point, the sales team should’ve been doing this for years—but it happens.
We’ve all been there.
Dixon Jones: Yeah, absolutely.
Check out the BuzzStream Podcast
