Andreas (Dre) Voniatis on Gaining Visibility in AI Search with Audience-Focused Content

Speaker A: Hi, everyone. Welcome back to another edition of Paris Talks Marketing. And my guest today is Andreas Voniatis, and he goes by the name Dre. Dre is the CEO of arteos and he's the author of a book called Data Driven SEO. And he's here to tell us all about the shift that is underway, the sand shifting beneath our feet in the SEO world right now as a lot of users are shifting to and between Google Search and large language models and AI overviews. So with that, Dre, welcome to the show.

Speaker B: Thank you, Paris. Thank you for having me. I, uh, look forward to sharing what I know about AI and LLMs.

Speaker A: Let's jump right into the deep end here. In your LinkedIn headline, it says, get recommended by AI. And that says seems to be the mission of artios of your consultancy. So what does being recommended by AI or AI recommended, what does that look like in practice for a marketer, like a CMO in 2025? And how is that different from ranking number one in the SEO terminology a few years ago?

Speaker B: What that looks like is that, uh, whenever your target buyers are looking for information, your brand is listed as one of the sources. Whenever your target buyer is searching for a partner or a vendor in what it is your company provides, you are one of the shortlisted brands.

Speaker A: Okay, and that is different than ranking number one in Google Search from a few years ago or even today, right?

Speaker B: Yes. I'm sure it will be of zero surprise to you. Many business owners and marketing directors are complaining about how all of the leads, traffic and revenue is drying up from Google. There's a lot of ads and AI overviews displacing the traditional web results. We're seeing AI mode being trialled in the US and no doubt which is like a ChatGPT style interface where there are no web results to be seen. And, um, that is only going to be rolled out across the world in the next three to six months, I'd say.

Speaker A: Yeah, I think this is probably the most major change that Google has made to search in at least 10 years. And I think it's the natural next step after AI overviews. Do you see AI mode as at some point replacing AI Overviews entirely, or is it something that will always be sitting there as an option for users to toggle?

Speaker B: I think it will replace the whole search experience and eventually we'll start seeing ads being shown alongside generated results. Just like Google started. It was just a simple search bar with organic results. Then suddenly you see a few ads, then, now we see lots of ads. And I think they're undergoing a reset in response to the challenge posed by ChatGPT and other LLMs.

Speaker A: I think their hand has finally been forced and they probably wouldn't have done this without the competitive pressure. But it's time to evolve and I, uh, guess they're really banking on the fact that they still have the users and they can guide and bring the massive user base slowly and gradually over to an AI chat experience and still preserve monetization. And right now it seems to me that they're sacrificing some portion of informational queries to the LLMs and to AI overviews. But the good news is that most of those are not very monetizable from an ad perspective. So I think that their ad revenue is still fairly well intact. But there's going to be some threat to that if they don't make this move soon.

Speaker B: Yeah, it's a really good point. I think eventually though, we will. I have no doubt that right now every second is being spent trying to work out how to monetize the generative results. I think search will still be available, but more as. More as a toggle, whereas AI mode will be the default.

Speaker A: Yeah, I think you're right. And the work that you do with your clients is about visibility in generative AI chats, in LLMs, and also AI overviews. And you told me that you're not trying to reverse engineer AI results the way that SEOs try to reverse engineer Google's ranking algorithm.

Speaker B: Yeah.

Speaker A: But rather that you're trying to reverse engineer audiences.

Speaker B: Yeah.

Speaker A: Can you describe what you mean by that?

Speaker B: Yeah. So reverse engineering search results is a bit of a, what I call a 10 Blue Links mindset from the traditional SEO era. And the problem with trying to reverse engineer the results is a bit like trying to, trying to write content, trying to rank AI written content or use AI written content to get yourself AI recommended. But that in itself is quite diluted. You wouldn't drink your own sweat to rehydrate. So why would you try and figure out how to be visible in AI or get recommended by AI by using diluted outputs? Whereas if you go to the source, like uh, for example, if you go to the fountain to drink your water or you. Or if you data mine online conversations or authoritative content that's being had, you're more likely to produce research that produces content that AI wants, because it's going to correlate extremely highly with the AI worldview on what it thinks, how a subject works, if that makes sense.

Speaker A: Yeah, I think, uh, the analogy that you referred to the last time we spoke was a cassette tape. For those of us who remember what cassette tapes were, if you made a copy of that cassette, you could make a copy of a cassette tape, the audio quality would be degraded slightly and then if you made a copy of that copy, it would be degraded again. And uh, at some point that's kind of what you're doing by using AI to create content for AI, for ranking in AI overviews or generative AI. So what practical tips, recommendations, uh, would you offer to our audience for people that still want to leverage the immense power of tools like Deep research, for example, to create a, ah, really, really well researched, thorough piece of content on a topic and maybe the best quality that exists on the web, but at the same time still include that authentic voice so that it doesn't get classified as a just pure AI slop, so to speak.

Speaker B: Yeah, sure. So the first rule is don't use AI to write your content. Second rule is, don't use AI to do your research because that's just diluted stuff. And uh, the reason why these rules serve you or the marketer is because if it's free for you, it's free for everybody else. So there is no competitive advantage to be gained by using diluted outputs of AI to do either. Now I get asked this question quite a bit. You know, clients, they go, love what you do. You know, it's really working. But I've got a neighbor who, or friend who's got a landscape business and there's no way they would be able to afford you. You know, is there something you can do that uh, is a, uh, sort of stripped down version of what you provide or deliver? And it's like, well, it's a bit like going into a Tesla dealership and saying, look, as a Tesla owner, I really love the car. It's a great drive, great driving experience. You know, I got a friend who wants the Tesla experience but can't, uh, afford a Tesla. How do I tell them how to build a Tesla? Do you see what I mean? Or maybe that's a bit of a harsh, brutal example. But another one might be how do you win a Formula one race or a motor race with a bicycle? So unless you're prepared to build the machine to data mine the Internet and reverse engineer your audiences, the next best thing that, that isn't building the machine or building a car would be to carry out surveys, like with Gallup, for example, because then you're getting real data on users, on what the target buyers are thinking so you can create the content that other target buyers can learn from, if that makes sense. And that uh, is how you would get yourself into AI, because it's a bit like this nightclub. There's only a fixed capacity, uh, or in AI speak, there's only a fixed amount of tokens or computing power and there's only a limited window for people to be, or companies to be surfaced in the AI results. So there's a very high bar and if you're not prepared to meet the high bar, you know, a bicycle strategy will not cut it.

Speaker A: Yeah, I think it's similar to advice that we've given in the past when the initial um, the instant answers started to appear in Google search results. And again that was a precursor to AI overviews really because there were very few clicks, they would give you the instant answer. It was great for mobile. And the rule of thumb for that was if you want it to be the answer, if you want it to be position zero, then you had to make a huge effort and you put a lot of resources into that. And it's kind of an all or nothing type of a game. That may be bad news for a lot of content marketers that in the last 10 or 15 years you could still build a decent business from average, uh, rankings of five to 10 or you know, if you're not in the top one, two or three, you could still get traffic. But do you think those people that would rank positions 5 to 10 in an SEO world won't have any visibility?

Speaker B: AI overviews world AI results as we've seen, is quoting anywhere up to 20 as much as 20 data sources. I'm not entirely convinced that there's not as much to play for. The other thing is as well is we're seeing a fragmentation or rather a multiplication of search results because search behavior has changed. In AI, people are a lot more contextual, they're a lot more specific about who they are, why they want it, what they want and in what format. Now before we had like a, uh, kind of one Yellow Pages for an entire national market or topic area, whereas AI has split that into many different regions, Personas. If anything, there's actually a lot more to play for. And if you want it all, it's going to cost you a lot more before you could get away with a, uh, Wikipedia style guide, you know, the ultimate guide to ransomware recovery. And you know, your article would satisfy students, MSPs, technologists, relatives looking to settle an argument over dinner, although I don't know why they'd be talking about ransomware recovery. Whereas now you could put together content reports on um, you now have to have a content report targeted at the us maybe by state. You know, the search opportunities are much wider. So I think this is actually a great opportunity for local businesses. The brand can't just march in with their one ultimate guide that suits the rest of the world. I think now they have to produce ultimate guides for, you know, multiple states, which is great. If anything, it's created opportunities. So I think the little guy can still get ahead.

Speaker A: You've said that the 10 blue links era, that mindset is a uh, hangover really. So what are the first one or two habits that now in house SEO team has to unlearn to survive in this new AI agentic search future?

Speaker B: I think one of the habits to unlearn is trying to turn um, their site into the Wikipedia of you know, low effort content. That is gone in my opinion. Also they have to be a lot more targeted towards when they do produce the content they need to be looking at making the content targeted to the buyer. Now, for example, just to take the ransomware recovery, having articles on, um, what is ransomware recovery is just simply, that has to be unlearned. What has to be learned is put yourself in the shoes of your target buyer. Your target buyer probably already knows what ransomware recovery is. So actually it's more important to have content a lot more targeted towards that target buyer. What is the most common ransomware attack? What were the challenges you had to overcome? How did you overcome it? What was the damage to your company? What was the cause? What things will you be doing in order to prevent them? There may be some overlap with a guide on what is ransomware recovery, but I don't think we're going to see those kind of guides succeed anymore. In fact, with AI now they're succeeding today because there's not enough people like us who have knowledge of what it takes to, to thrive in the AI era. So in the absence of those of the kind of content that we're producing, LLMs are forced to choose from the worst, if you know what I mean, or the hangover, uh, of SEO.

Speaker A: Yeah, I have an opinion on this, which is, I think the biggest hangover is going to be keyword research in terms of something that has to be unlearned. Because I think now our SEOs are still instinctually trained to start with keyword research. Every SEO workflow process, it does start with keyword research. I mean, not on the technical SEO side. That starts with a crawl of the website, but with content starts with keyword research. But I don't think that there is going to be a keyword research tool for AI and or engines. I think it's going to be more broad topic based and it won't go further than that. I just don't see because the ad models aren't going to monetize the same way on the basis of bidding for keywords. And who knows even what kind of AD revenue model ChatGPT will come up with. I think they will at some point. But Google Ads has to evolve into the AI mode and I think the AI mode is not going to be keyword driven or cost per click driven. So because of that I think keyword research as a primary uh, Skill set for SEOs, I think that's no longer going to be valuable. Hunting and finding keywords and then clustering them and putting them into groups, I think that's, that's gonna go away.

Speaker B: Uh, I think Paris, you must be reading my mind because I think all these keyword research tools like Semrush, I think they're dead, they're not fit for the AI era. Keywords hide who's actually doing the searching. So that's the first problem. You don't know who's actually searching these phrases. Also the keywords are all derived, the numbers are derived from, from Google Ads. So that number is designed to maximize the revenue per click to Google and nothing else. It's not there to serve the SEO or community or marketers. It's purely there to maximize Google revenue. So yeah, I mean we have our own infrastructure to infer the topics as you've said. And that is where the gold is. It's what is your target by discussing that is what now defines the AI content strategy going forward.

Speaker A: All right, so that leads us directly into a conversation around KPIs. And I think that if search impressions used to be the North Star KPI or actually that could be the North Star KPI in a zero click world. But in terms of KPIs, if we imagine that we, we're not going to be really looking at clicks and being able to attribute clicks and conversions back to a keyword. Uh, if that's all going away, what are going to be the major KPIs? Let's say I'm a CMO, I'm going to my CFO and I need to ask for budget and it's still going to be my performance marketing budget, which typically would go into paid search and the bottom of the funnel channels and I used to be able to show an ROI or a return on ad spend that would help me get that budget because I would say we invested 100, we got back 300 or whatever it was. But now attribution is going to get a lot harder. What are going to be the new KPIs that help marketers defend their budgets for performance marketing, especially when some of those budgets now need to shift into this, uh, new form of AI?

Speaker B: Yeah, it's a great question. You're absolutely correct. It's going to be really difficult to make the direct link between AI visibility and revenue or leads generated. But I'm sure we can all agree that there's always lead metrics, metrics that precede the desired outcome. And so I would say now it's kind of flipped on its head. I would say it's all brand searches. Because if your information is being delivered in generative responses, AI searches and things like that, then you're going to be increasing brand awareness and you're going to be increasing the number of brand searches. Now how long that will last is questionable. The other side of it is what I think has a much bigger future is tracking how many discussions are happening online regarding your brand. Because if that's increasing, then you know you're doing the right job. Then it becomes a question of multi channel and um, multi content path attribution modeling in order to tie those two phenomena up, you know.

Speaker A: Yeah. So does that mean that performance marketing itself is going to be going away?

Speaker B: I've never thought of organic search as a performance marketing channel. I've seen firsthand in some agencies how they tried to position it as a performance marketing channel and then they ended up siloing it or recategorizing it as a specialist service because there's too much of a lag in time. No matter how quick you are as an SEO, there's always a bit of a lag, which makes it difficult to put together some sort of, you know, there's no instant feedback loop like there is with Google Ads where your ad shows up, um, in, you know, within 15 minutes of sending it live. And that's near instant. With SEO and AI there is no instant feedback loop. So you are going to get a lot of disillusioned business owners or marketing managers in small companies that are going to be really disappointed because they haven't quite understood that, uh, organic marketing is not advertising.

Speaker A: Yeah, I think we're talking about going back full circle to brand building and brand first thinking. And with that, what sort of brand building Playbooks, do you think translate best into the signals that AI systems will respect?

Speaker B: You will still need to become a content authority in what it is you do or provide in the marketplace. Except the difference will be, you know, before it was just content guides to satisfy everybody. Now it's content guides that uh, are very targeted at your buyer and, um, by region.

Speaker A: Yeah. And in terms of the content quality, I mean this is going to demand much higher level of quality. And you referred to the proprietary value. Even outside of search, what does a high effort content, a high effort article really look like? In 2025? You mentioned doing original surveys and things like that. How do you test and validate that the content hits that quality bar before you publish it?

Speaker B: That's a great question. If you haven't already created an LLM like I have, you could try reverse engineering results and then we're getting back to that 10 blue links mindset. Again, generally speaking, like I, I've seen a lot of feature impact studies that uh, have tried to decipher what it takes to get AI visible and they all seem to fail quite miserably because they haven't included proprietary content and things like that. And how you measure that or simulate that beforehand is really difficult. So right now it's, it's the Wild West. It's really blunt. You would have to just go with, for example, one hypothesis might be what is a respectable minimum sample size? For example, Ernst and Young did a study on data subject access requests as a result of GDPR and they interviewed 500 data privacy officers in the UK. So right now you would have to go into simulation in order to start reverse engineering. But by the time you build the infrastructure for that, you could have just as easily got the research data you needed in order to get visible.

Speaker A: Yeah. And like the old adage of kids that are trying to cheat on a test and at some point the cheating method gets so elaborate that it actually takes longer than if you had just studied and learned the material. And I guess a lot of SEOs still are in a mindset that this is a game and that there are ways that you can win. And even going back to the black hat SEO days, there were tricks that you could play. And I think it's in our nature as SEOs to look for shortcuts, not necessarily doing things unethically, but to look for shortcuts to optimize our success. And I think it's going to be harder with LLMs. But can you help paint this picture? The risk Reward calculus for SEOs that are listening that might now be looking for a way to game LLM results the way they did with SEO.

Speaker B: There's a few hopes. Like I said, first of all, there's more to play for, there's more markets, there's more Personas to create content for. So that already reduces the cost. But there is no substitute for good hard research that tells the world that your content tells the world that the world didn't know before. So there's no getting around that. And if you try and cheat it, uh, with using AI to do your research for you, then just imagine statistically to an LLM, how you know me too or Samey, your content looks compared to everybody else. There's a lot of competent SEOs out there and we all like to think we're the best or one of the best, but really the reality is that, uh, there's a lot of shared knowledge online and there's always someone who knows more than you and we don't truly know everything as much that fantasy is strong in our minds as it is

Speaker A: a harsh truth there. Yeah, I think in my view it's going to be more and more about the deeper and deeper knowledge of your Persona who you're selling to because the LLMs will also know those Personas really well. I've been shocked so far at how much ChatGPT already knows about me. And in one conversation it will reference something that it knows about me that I thought was out of context of that conversation, but it would find a way to make a connection and by doing so it would remind me how much it really does know about me. And I think uh, Google also knows so much about us from all of our online activity. And so maybe it's not necessarily the advertising or being mentioned and included in these answers may not always be strictly tied to the context of that conversation happening in real time, but it could also go back to some other memory that the LLM knows about you. I've turned on memory in my chatgpt, but I think it was there already because even earlier, before I had turned it on, it was still referencing things. So to give you an example, I have some arthritis in my knee and, and I've been managing for years now and I might ask a question, a uh, health related question about what's the impact of eating this type of food or something like that and it will actually bring into that answer that this might help reduce some, some inflammation in your joints, it might help with the arthritis and that that could be an opportunity to plug Maybe an advertiser or a content provider related to arthritis, but not necessarily related to the, a recipe or food that I was inquiring about. So it's because of the deep knowledge of me as a Persona and the whole richness of my uh, all of my needs. I might respond well to a piece of content, whether it's organic or paid. As a user I shouldn't care. But if I'm reminded that, yeah, this is something that could help with arthritis, and here's another tip related to that, I might still go down that path. I might actually want to investigate the source of that. I think in that case marketers are going to turn over a lot of the context to the LLMs and we just have to really be available as much as possible. I think. And understanding um, that it's not just a particular intent in one moment that someone has, but it's a person who is our Persona that we want to reach and we'd like to get in front of them at any moment where there's some connection to what they're talking about. Am I making sense?

Speaker B: Yeah, I think two thoughts come out of that, uh, when you talk about uh, the personalized knowledge. I think it's going to be really important for companies to get ahead of AI because right now target buyers or your audiences are uh, prompting. But uh, the future will be much more proactive and agentic. So right now it's pretty reactive. In the future it will be more proactive. You'll have agents that already know more about you. But I think what we'll probably do is we'll have like some sort of document that we kind of upload and our uh, personal agent, AI agent will know not only who we are, but it'll know what we want, when we want it, how we want it without us having to prompt. And it will do all that work in the background and it will give us a heads up when it finds what it is we were looking for at the right time. And so that would deal with things, health issues or challenges just as like the one you mentioned by example. And I think the other thing is it may force the hand of publishers to do more than just text content, but actually do things like apps and things like that to offer a, uh, much more interactive experience in order to get ahead. And we may see an evolution in the sort of content contract between publishers and AI LLMs where APIs may be provided to LLMs so that LLMs can serve the app directly within their results and reward the publishers financially for it.

Speaker A: Yeah, I think some of those publisher licensing agreements are starting to be made. Those deals are starting to be made, but it's early. Otherwise, what else is the incentive for anyone to maintain a website if that website is just chicken feed for large language models and people don't visit them anymore?

Speaker B: Well, there are other channels other than AI, so you still have your social. If you're B2B, then it'll be intended mainly. It's one of the Ps of the marketing tactics. You know, place. It's where the sale takes place, space, and it still gives you information. But again, I don't think it's entirely accurate to describe it as chicken feed, because AI has now set the bar higher, which means that one content can't rule all. You have to have content specific to your Persona and to your market. And, you know, if you're doing a report on 2025, you'll need to inevitably do a report in 2026 if you want to stay relevant. So there's already three dimensions to multiply the opportunities unlocked by AI search, which we had a very limited interface with Google because it seemed that, um, one ultimate guide would rank for multiple years, even though it was out of date. It would serve every region in the world and never mind an entire country. And it would appeal to every Persona, you know, from students to grandma, your target buyer. So I think AI has actually done us a favor. It just means that, uh, you can't be greedy and expect one content report to give you the world's custom or business. You need to work that little bit harder.

Speaker A: Yeah, repurpose that stuff.

Speaker B: Yeah, 100%. It's like the Yellow Pages. One Yellow Pages for one country has now gone into several states or counties within states.

Speaker A: You know, Dre, this has been a fascinating conversation and a great time to. To be in marketing. It could be a little bit unnerving for a lot of us. A lot of people are seeing organic traffic that is. Is dropping, but maybe at the same time, brand visibility might be increasing. I think we are slowly coming full circle back to the essence of marketing. The future is going to be less quantifiable in terms of clicks and keywords, but I still think marketing is going to be as necessary and as relevant as ever. People still need to find a way to learn about products and services around them, and it's going to be a great adaptation ahead of us.

Speaker B: Can I just say that if you're producing content that's winning in AI, unlike SEO, you can get surfaced in AI pretty quickly. Like within one month of your content going live for highly competitive searches. And that's how it was in SEO 25 years ago when I started. Right, except the window for AI for the small business is going to be much shorter because everybody today knows what SEO can do for a business. Whereas 20, 25 years ago a local business that uh, I was helping for SEO, you know, managed to multiply the size of their company because nobody knew what SEO could do for their business. And the window was much longer because the giants were fast asleep, leaving SEO to the graphic designer or the site programmer who was not that much wiser about how SEO or search engines work. So the window today is much shorter. But if you're winning in AI, it's not just AI. If your content is good enough to be valued by AI, it's good enough to be valued in social video newsletters, webinars, downloadables. So there's all to play for. But yes, there's a higher bar in terms of the content quality, but there's also higher reward. SEO content playbooks were only good for search, but AI content playbooks are ah, good for everything.

Speaker A: That's a very good point. The higher stakes come higher rewards for sure.

Speaker B: I mean the AI content we're producing is so thought provoking, it gets discussed on LinkedIn. And even rival law firms are linking to my clients content indiscriminately. And that's what I'm saying. And this is Another thing that SEOs need to unlearn is that uh, this is value beyond search. And there are even more opportunities in search than there were before. AIs multiplied those opportunities because of those time Persona geo dimensions. But also now it's multi channel. You know, it's uh, like if you've got content that works for AI, you've got content that's great for all of these other channels. So uh, I'm seeing greed and opportunity here. I'm not feeling that fear and certainly my clients aren't.

Speaker A: It's tempting to be fearful. Clicks are the thing that are monetizable because there's a click, there's a visit and then I have a conversion rate and then I have some revenue. But that formula that's driven by clicks and visits might be going away, but there's still uh, many other pathways to win business. And as you said, if you're going to build content that is optimized for AI, then it's going to have great reuse value in a lot of other channels as well. That's a very good point. All right, great. Thanks for spending the time. And where is the best place for people to find you? Online?

Speaker B: Thank you for asking. I'm on LinkedIn and my website is rtos IO a r t I o s IO.

Speaker A: Great. So for those who are listening and want to meet with Dre and understand how he gets his clients brands inserted into these conversations in ChatGPT and AI overviews, check out RTOS and find Dre on LinkedIn. And thank you for spending the time with me. It's been a great conversation.

Speaker B: Trey, thank you for having me. It's been great to speak with like minds.

Speaker A: Absolutely. All right, take care.

Speaker B: Cheers.