Webinar: AI in the life of an SEO Professional

Remember when you chose an SEO career because you loved strategy and creativity?

Unfortunately, a big chunk of SEO ended up as endless repetitive tasks eating away your precious time while clients expect more and competitors race ahead.

The most valuable SEO skill today isn’t technical knowledge – it’s knowing which tasks you shouldn’t be doing at all.

Fancy learning how AI can handle your mundane workload while you focus on game-changing strategy? Our webinar showcases real-world examples that will transform how you work.

See also: AI prompts webinar

Full Transcript of the AI in the life of an SEO professional Webinar

Peter Mead:
It’s Peter Mead and welcome to this Duda webinar. This special topic that everyone’s talking about and everyone’s trying to make sense of in this whole new world that’s emerging—SEO, AI, what are we going to do? So today, the topic is AI in the daily life of a professional SEO. So I thought, who better to get together than actual professional SEOs and talk about what do we do on a daily basis with AI.

So, first of all, I’m going to go around and introduce everybody and read a little bit about each person. Today we have Sally Mills. Sally Mills is the Senior SEO Lead at InfoChoice Group and a driving force behind Brisbane SEO networking, based in Australia. She brings a hands-on, practical approach to SEO, combining content strategy and technical implementation. As a community organizer and a regular speaker, Sally is passionate about knowledge sharing and helping others to stay on the cutting edge of search trends, especially when it comes to AI. Sally, you know, I’ve known you for quite a while. You just recently have also had an SEO meet up for the SEO Collective, and you had a big presentation about AI. Do you want to add some more? Welcome to the show! Do you want to add a bit more?

Sally Mills:
Yeah, this is really well timed because last Thursday, SEO Collective in Brisbane, which is a free event that we have in Brisbane—there’s also ones in Sydney and Melbourne as well—we had a big panel together and all talked AI and SEO. Because as we all know, our roles are all really changing and it’s just ramping up more with AI, which is exciting. So it’s never boring in SEO. Yeah, never boring.

Peter Mead:
Thank you so much. We’ll be so keen to talk with you about some of your opinions during this webinar. But let’s first of all say hello to Natalia Witczyk. Natalia is an international SEO consultant currently based in Tianjin, China, originally from Poland. She has over 10 years of experience in the SEO industry, gained from working with agencies in the UK, later in Spain, and as of this year in China. Natalia specializes in technical and international SEO, developing strategies for large international websites across various search engines, including Google, Yandex, Baidu. She’s also the founder of Mosquita Digital, an international SEO consultancy, and as an SEO specialist at Jade Demon Digital and active member of the SEO community, Natalia is a frequent speaker at industry conferences and serves as a judge for search awards. And also, Natalia, I’ve noticed you’re going to be talking at the upcoming APAC Google talks in Bangkok.

Natalia Witczyk:
That’s correct. I actually haven’t announced it yet on my own LinkedIn, and I see Peter already has done his homework and found out I’ll be speaking at the Google event, which I really can’t wait for. And yeah, hello everyone. Thank you for having me.

Peter Mead:
Thank you so much. And with all your experience, I’m very interested in what you have to say today. So next, we have Fabrizio Ballarini, Head of Organic Growth at Wise, who leads organic growth SEO at Wise. It’s a financial company powering money for people and businesses in their increasingly global lives to pay, to get paid, to spend in any currency wherever you are, whether you’re doing so for a team. The mission at Wise is to acquire customers sustainably by building products, platforms, and content that educate and solve customer problems, working across SEO, affiliate, CRM, and social media. I personally witnessed one of your original content marketing campaigns, you know, several, quite a few years ago now, Fabrizio, and ever since then I’ve been very interested in what you’ve been getting up to. So thank you for joining us.

Fabrizio Ballarini:
Thank you very much. Yeah, it’s great to join. We, as you know, you guys are on the other side of the world compared to London, but actually our team has been expanding to Asia Pacific quite a lot, so I spend quite a bit of time in Singapore and working with people on that side of the world, so it’s exciting.

Peter Mead:
There’s a lot to get through, and our terrific guests have very kindly spent a little bit of time and created a few short slides. This is not meant to be a huge big tutorial, but it’s just a few slides to give some visual explanations of some of these topics.

My personal opinions here—and I think maybe what I’m experiencing, what I’m hearing from SEOs, maybe SEOs that have been in the industry for a long time—you know, I mean, SEO really always has been thought of as, you know, SEO is traffic, it’s all about getting traffic to your website. And of course, unless anybody’s been living under a rock, that whole kind of value proposition about what is SEO is changing, right? It’s really changing. Of course, we now have all these things that have happened. Several years ago, ChatGPT 3.5 became really useful and helpful and started challenging Google, and Bing came along and started offering a bit more, and of course now we have AI mode, which essentially is a Gemini prompt on the screen, you know? So, I’m just really interested. Maybe I’ll just go around and ask each person.

Sally, how did we get here? I mean, are we still doing SEO? I mean, some people are calling it GEO, some people are—what are we doing? Is it SEO as we know it, or is it not anymore?

Sally Mills:
I think the moment there is a search bar on the page, it is SEO, whether it’s on Google, whether it’s in GPT, whether it’s in social media such as TikTok. So, I don’t think SEO is ever going to be dead, as they like to say. I do think it’s an exciting time in that it’s just changing, right? Like, we’re very much getting a new cycle of what we’re optimizing for. I remember when featured snippets came out and we were optimizing for featured snippets, right? And it was all the “Oh, how can we rank?” and “Oh, you test different things,” like putting an H2 and then answering in 40 to 50 words, and then, “Ah, I’ve got the featured snippet,” right? Whereas now we’re all back to the drawing board again, going, “How can we have our brands appearing more in AI instead?” So, it’s just this constant, ever-changing version of SEO. So, we’re not dead yet.

Peter Mead:
Yeah. Sally, you said that you have prepared a few slides. Would you like to talk us through for maybe the next five minutes or so, just talk us through a little bit that you have there?

Sally Mills:
Yeah, let’s do some quick stuff, because I don’t have any sort of prompts for you today because I know everyone’s always got their hard-hitting prompts. But I think the biggest thing for us SEOs is sort of getting our brains thinking and making us faster at our jobs. Two years back, I spoke in Sydney about sort of automation, so I’d use another example from there today. But what I wanted us to just quickly go, because I had a think of what do I use AI for in that sort of day-to-day thing, and what do I also dream of doing, because there’s some stuff I haven’t yet done. And I’ve sort of broken it down in that technical area to three different things.

So, there’s sort of that Google Colab Python stuff, so it can take us all from that level of being not coders—and I’m an ex-web developer, so I’m pretty lucky that I’ve got that little step up with a bit of code—but it gives you that ability to go, “I don’t always need a tool to do things.” The classic example is redirect mapping. We can now use Python with the help of GPT to teach you how to do the scripts or even write the whole scripts for you—vibe coding, I suppose—to do all your redirect mapping for you, and then you’re just really a QA. So, there’s all these sort of manual things we used to be doing that now we can really automate, and where our jobs are much more QAing in a sense.

Things like keyword clustering as well, so you can go and use scripts with Python with the help of GPT to go and actually cluster your keywords rather than manually sorting like we used to back in the day to work out these topics and what relates. Internal link mapping, even image renaming—like, I remember using GPT just something really ad hoc, right? Like, “Oh, I have all these image files that are all named a bit weirdly, I want to strip out—say there’s too many underscores or something in the URL,” you can use GPT to write you the script to then clean them up for you. So, there’s just so many different things that just levels us up as SEOs to being pretty good, just average, below-average web developers. So, it’s really, really cool there.

The next one is obviously everyone here would have been just doing these sort of ad hoc ones, whether it’s your regex problems. I actually am a bit of a nerd and I really love regex problems, so I’m almost a bit sad that I’m losing my regex problems because I used to sit there and be like, “Oh, we can definitely do this, like, we can do this redirect with one regex, right? I’d solve it.” But now I just go to GPT and GPT does it for me, which is sad but good.

Things like JSON schema—I’m sure everyone here has been using GPT for, or whatever AI model you’re wanting to use for JSON schema—sheets formulas, of course, and then HTML, JavaScript, CSS. We’re all developers now, which is just cool, because we’re just able to move at such a faster pace as SEOs.

The next one I put here was Screaming Frog. I’m a big fan of them, and I was reading their release, and they did another release again, and I’m struggling to keep up with all the cool things they’re building in. But it just sort of tugged my memory that it’s not always going to be just us using AI, right? It’s actually the tools that we’re using are upgrading again and again and again. So, there’s plenty of things you can be doing—Screaming Frog with an OpenAI token and everything. So, you could be doing image generation, so if you had a whole heap of blogs that needed banner images, you could go and just run it through Screaming Frog and it can generate off your prompt as you go. Semantic content analysis, so that would be looking at whether things are similar or duplicate, which is just so helpful for—I’m pretty keen to give that one a spin because I work on quite very big websites, so the ability for it to go and decide what’s quite duplicate, whether we need to consolidate content, is just so good. I wish I would have had that like two years ago when I was doing big migrations and cleanups.

And then even just things like doing your meta titles, descriptions, headings. I saw you could do like product page content. There’s just so many things you can be plugging into the tools we have access today. So, there’s a few examples there.

Just an old example I used to use—the script’s actually on my website, so you can all feel free to go and steal it—it’s just being that fuzzy match redirect example. So, you’d be able to take a list of old URLs that need to go to new URLs, so you’ve pulled all, maybe it’s all your 404s from a Screaming Frog or maybe you’ve exported from Search Console, or maybe you actually just need to do new redirect mapping. And you can take both lists of these URLs, rather than us sitting through and going and matching things, especially if you’ve worked on really big old websites, there’s a lot of history in them, you can instead just be using basically fuzzy match and it’s called Google Colab to go and do it for you. So, it’ll go and look at everything, it’ll run it through a model, and then what it will output is a bit of a similarity match. So, that’ll tell you that it’ll do a lot of work for you, because if you’re looking at everything that’s, say, 80% upwards, sweet, let’s put those redirects in, job done there. If you’re looking at—it’s only when you get down towards those smaller similarities that you’re then going to have to actually do a bit of manual sorting, because we’re still having to check things and we’re always going to be checking things with AI.

Yeah, so that’s my quick thing to hopefully get you all thinking about all those little ad hoc technical things and leveling up.

Peter Mead:
Ah, yeah, that’s great, Sally. Interesting. I mean, it seems like a real—such an obvious thing when you’re talking about getting the AI to do things like page titles, meta descriptions. It seems a little bit like, yeah, this is a no-brainer, let’s just get AI to do it. But I’ve got a little prompt, I might just share it in the chat here. I’ll share it in the chat. I’m not sure if—Anton, can we put that up? Can we get that put out into the chat, that little prompt?

I think the idea here is that there’s a prompt here. This is—the team’s using it. One of my guys, Mark Whitley, he’s a fantastic SEO, works with me, works for me in the team. And while I was looking, I just find this—I mean, I’m a little bit like where you said you were sad to—you know, you like doing regular expressions and a little bit sad, that kind of thing. I’m a little bit far beyond being sad. I’m kind of a bit jaded, really, about sort of AI. Look, it’s one of those things, we need it, it has to happen, you have to use it every day. Do I like it? Not really. You know, I’m mean to it, you know, like, threatening to never use the AI again. I’m not, you know, like, “This is what we have to do to get a good answer out of you. Like, please give me some good meta descriptions. If you don’t give me some good answers, we’re just never going to use you again, like, you’re out of a job.”

Sally Mills:
Yeah, it’s really still struggling with that sort of character count, hey? Like, sometimes it will finally do it right and other times it won’t. I’ve recently found that if you’re using the 03 model on GPT, it’s a lot better. Makes it stop and think compared to four. I feel like it’s still hallucinating a bit sometimes.

Peter Mead:
Okay, yes, very good point. You’re touching on different models. Certainly, there’s a whole can of worms we could go down there about all the different models.

Let me just go across to Natalia. Natalia, you’ve got some other things that you have prepared. Maybe share your slides and talk us through that, and it could spark off a bit more conversation for us all here.

Natalia Witczyk:
To start with, I just want to say that I do use AI a lot, but I’m a skeptic at the same time. I’m not full on “move on to AI everything you do” kind of person, not at all. And actually, just this week, my LinkedIn feed has been full of people complaining about the fact that OpenAI just blocked their account because maybe doing HK project, some word or some phrase that’s deemed forbidden or blacklisted or whatever was there for a reason—we don’t know. They don’t even know which word that was, and they cannot use any of their custom GPTs, their whole workflow is broken. And I was reading that and thinking, I’ll never be that person because I’ll never outsource that much that I cannot just go without it at all, right? So, I just want to warn everyone, don’t put all your faith and all your eggs in one basket being OpenAI, that can just kick you out anytime they want, because it’s a privately owned company that has no rules to follow. If they don’t want you there, you’re banned.

Starting with a warning, let’s talk about AI in a more positive way.

Peter Mead:
Thank you for helping our audience to think of that, Natalia, because I think it’s very important to remember that these tools are third party tools, right? And I think you don’t think about it until it happens to you, right? And then you’re like, “I wish I’d known.” So, yeah, now you can’t say that you didn’t know. Thank you. All right, let’s have a look at what you’ve got on your slides.

Natalia Witczyk:
I’ve prepared a quick few slides on how I enhanced my keyword research process, and I modestly said, “not the way you think,” because my approach to keyword research is not actually finding keywords through AI. That’s not at all what I’m doing.

Let me quickly show you, first, starting with what my B2B keyword research process used to be prior to AI, so you actually understand where I’m coming from first. My B2B keywords research was quite a specific process because the classic SEO part is very, very top of the mountain I’m building up with my clients, and I start that mountain off with in-person interviews. So, I used to run those in-person sessions, typically in the client’s offices if possible, by location, to have direct interaction and get the insights about the business from them.

Why is it relevant for B2B? Just because with B2B clients—and those of you who have that experience will probably agree with me—it’s a lot about nuance on what they do, how they do it, what they sell, who they are trying to target, and it’s not always as approachable as B2C brands, which you just get because you’re one of the consumers typically, right? So here, it’s really important to make it really specific while also understanding the wider keywords you can use, so it’s a bit of both worlds.

So, yeah, in-person interviews. And then that interview was always based on design thinking methods, so I tried to be as creative with questions I asked, and I would come with prepared drawings and drafts of things I want them to map out. So, there would be a lot of manual labor before we actually got to the session itself, just to make sure I get all the information in an organized but also creative way.

And then, that was always followed by the right questions. So, my favorite starter question would be, “You’re sitting at a Christmas table and your uncle is asking, ‘So, what is it that you do, and what is your industry exactly like, and what problems do you solve?’ And again, explain it to me.” And that’s getting those clients to actually numb the language down and throw out the jargon, right? So, I would ask a lot of right questions.

And then the hardest part was taking the notes, because I was there noting everything down and making sure every word that can be relevant later, I write it down, I pin it down, and I connect it correctly, and I understand how to map it. And that has always been finalized by the classic keyword research, when I spreadsheet it out, map the keywords to pages, suggest new content, suggest to merge content, but that was a lot of manual labor to actually pin it all down and map it in the proper way. So, that was my initial process that was very, very labor intense.

Now, this is my AI-enhanced keyword research process. So, there’s no steps anymore, because many of those elements happen at the same time. It’s just a process that keeps complementing itself. So, first, I have human-to-human interaction. And I said interaction because it’s not face-to-face meeting anymore, it’s actually done online now, and I’ve added a small “s” to indicate it’s many interactions, because now, because of the resources I can save thanks to AI, I can have multiple meetings one-on-one instead of a huge meeting for the whole team.

Then there’s human-to-AI interaction, which is the note-taking that I don’t have to do anymore. So, in the moment, AI is interacting with the humans on the same call while I’m just talking, AI is taking the notes. I use—in a second, I’ll show which tools—but I transcribe the actual conversation, not just take out the AI summary. I actually want to see what was said at what point, which words were used.

And then there’s AI-to-AI interaction, because whatever happened in the call, whatever was not taken, then it can be mapped out and clustered as shown before. This is the one mastering comes in hand, but obviously the initial part has been done as well.

The free tools I actually use here: Google Meet, just because it’s easier, I found, to integrate with any note-taking; Read AI, I use for note-taking, transcription, and also for the summaries; and then Gemini to organize the data, to cluster it, to give me Google Sheets already organized with all the information I need the way I want it.

And just to give you a few benefits, I’m saving so much time. I’m saving so much time that I actually can have individual meetings instead of one with the entire team, so I get better perspective, a wider vocab, just because every single person from the same team will tell me the same thing differently. So, it used to be really tricky when you have a full team of five, six people telling you one by one what they do, they would really replicate what had just been said because they would have been influenced by the colleague. Now, it’s separate meetings, and I actually have a chance to interact with more senior stakeholders, because I can speak to very technical people who typically would have been too busy to just join a company-wide meeting, but they can jump on a 30-minute call.

I also find that gives me more access to the company information and also gives me more seniority in my role as a consultant, because I speak to maybe head of sales or head of technical development, whatever it is, and they see SEO as part of the conversation. Those people typically love to explain what they do, and they really are interested in telling you really well what the company is about. So, this is a great thing for me to actually introduce myself to senior people.

Now, enhanced presence—because I’m focusing 100% on what’s said instead of taking notes, I can actually ask better follow-up questions, I can be more present in the conversation, and I can squeeze out way more from a one-hour meeting than I used to.

And quality outcomes, because as we all know, and I think we will say it a lot today, better quality input to the AI, the better quality output. And my quality is top-notch because it’s a human interaction input, it’s subject matter experts I’m speaking to, so whatever reservations anyone can have with AI here, they disappear because I’m inputting the human part, the highest quality human part. So, my outputs are really high quality as well, so I don’t have to worry about quality. I obviously also review everything myself, so I know exactly the outputs. I’m never just delegating to AI, I’m helping AI give me more time to do what I do best. So, that’s my keyword research process.

Peter Mead:
Wow, thank you so much. It’s very interesting how you’re saying that you used to have to have the whole team or a room full of people. I mean, part of my approach when I’m talking with clients is to do a similar thing, where if we’re talking about some new content, is to really just talk that through with the client, talk the subject through, and record that. It’s interesting to see you’ve got a very organized way of going about it.

Natalia Witczyk:
Yeah, organized and also, let’s say, pressure free. Because before, I was so stressed just taking the notes and asking, and now I can actually dive in and let the conversation flow, so I really enjoy the process more as well.

Peter Mead:
What’s the feedback from the clients been? Do you think you’re getting better feedback now from clients?

Natalia Witczyk:
I’m getting fantastic feedback so far, and actually one of the clients I did that with already posted several LinkedIn posts because what we recorded were such a great anecdote of his business and what he does. He told me he struggles with finding good LinkedIn content that is actually approachable, so I took parts of the transcript, smoothed them out just slightly with Gemini to make them sound more like written text than a set said text, and he just took it and he was so happy with it because they were his words, there was his story, and he just went on and posted it on LinkedIn as well. So, I think I might really expand my services to extra LinkedIn consultancy on the side.

Fabrizio Ballarini:
I think that definitely anything that is repackaging, cross-platform, will be on steroids at some point. Because even like clipping video, even like doing stuff that in the past you would have almost taken—like, you would have, back in time, considered, “Is this a new role that we have in the company? Or maybe we just don’t do it.” And probably many people just avoided doing it because maybe the value was not great yet with the content that existed, right? It’s not that it didn’t exist, it’s just it was hard to repackage it very easily.

Peter Mead:
Yeah, and okay, so just while we’re talking about some of your approach there, Fabrizio, I know you’ve prepared some slides and I’ve seen some of what you’re going to talk about. So maybe share your slides and talk us through.

Fabrizio Ballarini:
Yeah, sure. Back to the question that you asked earlier on what do we do and is it completely changing. I think, at least in our organization, we have to be pragmatic about what we’re investing in. Our finance team doesn’t necessarily care whether there is AI Overviews or not—they know that they give us some budget to invest in growth and then we return that investment, right? So even topics like drops of traffic and so on, it doesn’t mean that activities need to be deleted. We didn’t fire the entire team and start to work only on AI Overviews. We continue to do what we do at scale, but it’s hard to not acknowledge that things got significantly better, to the point that a lot of processes and activities will have to be different, both for us who do SEO and create content for our customers, but also for the users who use all these chat assistants as opposed to searching with AI Overviews.

This is just an example between two, on how our video got better. On our side, we are breaking down in two areas how we think about it. One is what we call automation—how do we do what we have been doing? This is the part of “don’t worry, nothing is over, we continue to do it until financially what we do makes sense.” This part relies pretty much on the same principles we’ve been operating on for years with regards to SEO. It’s possible that some areas have or will lose some traffic, but you always have to look at what the return on this is.

The second part is the optimization part, where you say, “Okay, because user behavior is changing on platforms, what do we do to increase our visibility in these new ways that users are searching?” On this second part, that’s where we are not behind—we are monitoring what’s going on quite heavily, but we’re not taking action yet, just because we don’t want to make up things to do that we’re not so sure about. Still, what we see value in right now is to continue to do what we were doing, but doing it extremely faster, with a bit more quality over time, and eventually reshape the process.

Back to Natalia’s point as well, even at the end of the slide, we have not reduced any—well, we have a pretty large team, we operate globally on multiple markets, and we have not reduced the effort that our team puts in. We just gave them a lot of powers that they didn’t have before.

Just to jump into it, the part where I think we can share the most today on what’s going on on our end is how we think about giving our team models so that they can be more effective. The first level is quite simple: us providing to our team the equivalent of Gemini and GPT—they can choose whatever models they prefer to use as a chat assistant. The second level is to have some sort of copilots, where some of these assistants within the platform that we use have got pre-input instructions so that they tend to be more on point when stuff happens.

The third level is where we basically had to form a team. One change compared to before is that now we have a team of a few people plus engineers that work on building agents. The main difference between what we call agent and the previous levels is that one and two is just one model; agents are basically multiple models for multiple tasks, plus things like scraping and other stuff that wouldn’t generally fit within what the models do. The last one is what we call network, which is basically all these agents talking to each other.

Just to make it a little bit more specific: the first level, you just chat to a model. The second level, you chat to a model that is trained on some wikis and documentation that we provide—things like our tone of voice, our compliance guidelines, and so on. The third level is that the model can extract data from URLs, can scrape, can do a bunch of things, thanks to us building an agent. The last one is where these agents talk to each other, so we go almost end-to-end in terms of doing our activities. For example, in our content creation process, we produce a lot of content for our customers, and we have built agents for each step. Again, it’s not that the person doing that task is gone; that person is doing it with the help of an agent, and their role has shifted a little bit, but they still cover pretty much what they used to do.

There are some tasks—also being in a regulated company—that maybe we didn’t enjoy as much as SEOs, like doing compliance checks and legal checks. The team is very happy that part of this exercise is now helped by the agents as well. So in that way, we go from an initial problem to eventually even editing a page with the help of the assistant. The ultimate step is integrating this into our documents, CMS, databases, so that the person operating this doesn’t only get the answer and then has to do a bunch of copy-paste, but whatever these agents are doing can eventually make changes either straight into the CMS—we’re getting there with some of the platforms we have—or at least to docs, or with certain markdown formats.

Basically, think about the content creation process: we used to do key research, briefing, writing and proofreading, compliance review, conversion to markdown, and publishing. Each of these processes had some steps and guidelines to cover. Each of these activities was taking some time or multiple people involved. It still takes time, it still takes people, but we are definitely faster and, I would say, even more accurate. Even things like key research—we have 40-plus specialists producing content in 15–20 languages, we produce 300–400 articles a quarter in all these languages. There must have been someone that did already the key research for the previous person at some point. This is the kind of stuff we try to avoid now. It’s not that we completely went full automation onto it, but at least we eliminate wasting time on stuff that in the past we could have almost not avoided.

That’s it. We have not replaced our specialists—the people that work on SEO at Wise are still in the team, actually growing the team. Wise.com is growing, but they have a lot of superpowers they didn’t have before. On average, this is good. We had the situation where people were skeptical. For instance, our first iteration of giving people an agent to do a content brief for an article had almost 55% negative feedback from the team—people were like, “No, no, no, I’m going to keep doing it the old way, it’s actually more efficient.” The more we iterate, though, and the more we get feedback, now on some activities the feedback from the team is quite positive. It’s not that every agent works with the magic of just pressing one click, but I think we find inefficiencies over time, and there are many tasks—and this is not just SEO, right? This team is now helping to build agents for other teams within marketing. What we do in SEO is quite specific, but also when we do email marketing, when we create content with our influencers or affiliates, there’s a lot of tasks that in the past used to take a bunch of spreadsheets and crunching, and these days we can automate pretty much.

To be honest, it’s a little bit like what was happening when I first joined my first job in an advertising agency at Google. There were people from the creative team talking about us as performance marketing people being “spreadsheet marketers” just because they didn’t like to use spreadsheets and thought it would have been fine without. There will probably be some people that will class some of the marketers as “AI marketers” also because they initially don’t like it so much. But I think if something is useful, people will use it, and that’s what’s happening.

The thing that we don’t know—and that’s why, differently from some people, I refrain from inventing a new name for SEO or trying to give a new checklist—is that we still don’t know, at least in my view, what is the adoption of the user on the search engine. My father doesn’t use ChatGPT, my father doesn’t understand the difference between an AI view and clicking on the template link. This part is a little bit more unknown, and in some areas it’s true that some websites have seen a massive loss in clicks to the website. Funny enough, sometimes when we look at scale and on a larger portfolio of domains that we manage, sometimes you initially think it’s the overview, then you go deeper and find out that one section of the website lost a ton of traffic for some other reason—the good old reasons. People need to be specific on where, but probably it’s true. We’ve seen this in some of the affiliate partners, we’ve seen it in some places where certain types of websites have gone a bit south compared to what it used to be, at least from a click perspective.

The other part is measurement. In social, we’ve seen a lot—if you do Instagram or TikTok, don’t expect that a lot of people will click out, and you have to get used to it, and it’s still a good idea.

Peter Mead:
Oh wow, that’s very interesting. A couple of things I just wanted to maybe ask you some more about from your presentation. I found it very interesting where you said the role of the SEO specialist has been given “superpowers.” So in some ways, they’ve actually got more work to do, not less. Is this the right way of thinking of it? How are you thinking of it there, in terms of the actual time it takes people to do tasks? There’s a lot of fear around AI taking jobs away.

Fabrizio Ballarini:
It’s possible that some tasks will not be as necessary. Maybe some tasks were not so valuable even before. In the past, to categorize 10 million keywords, you would have paid someone to do it pretty cheaply for a few months. Maybe that job is gone. Or otherwise, when we do outreach—not for links, but for influencers—we used to have people externally categorizing target lists of people to outreach and understanding the quality of some of these people. I think some stuff is fair to say that there will not be humans doing it anymore. But then there’s other stuff that we do. Even the part of QA, we see this happening in the localization team. As part of this group, we also run a localization team—the people who translate Wise in all the languages we are available in. Compared to many years ago, 90% of their work is machine translation, translation memory, and LLMs. They only translate 10%. None of them got fired because of it, because we still have to make sure that we are compliant 100%. We need to do QA. Even though AI is as smart as someone translating, we still need to have someone check that it’s a good idea to say something in this country. Especially, we have responsibility to our customers—our customers could lose money, there’s a lot of things that could go wrong. So I think probably there’s more QA being done than before. That’s what we say, but to date we have not been in any situation where someone’s job changed to the point that we had to change plans.

Peter Mead:
I think that’s fantastic. There is a lot of negative, maybe fear, maybe people are talking and saying AI is good enough. Let me just go back—so search engine optimization, SEO, right? That acronym. I mean, some people, you can start playing with the acronym and call it GEO or even say SEO stands for “search everywhere optimization” or “omnipresence” or whatever, all these different things. It’s funny, though, because I’ve been doing this kind of thing for quite a while, and a lot of the fundamental things that everyone’s talking about nowadays—it’s actually not that different to all the fundamental things that I’ve been doing for a long, long time. I just find it interesting that all of a sudden now, AI’s come along and—I’ve just found that the transition, actually, because if you’ve already been on a good program of working with your entities, your semantics, your EAT, all of the sort of things that would have got you good results before, AI mode seems to be the same thing which is getting you reasonable results in AI mode, including all that offsite, that third-party placements. So, I don’t know, do you have any ideas, Natalia? Do you think that we should change the name of it, or if someone says we’re doing GEO, does it make any difference?

Natalia Witczyk:
Well, I specialize in international SEO, so when you say GEO, I’m thinking about location, localization term. I’m never thinking about another discipline. I’m not completely opposed to expanding the meaning behind SEO. I don’t mind SEO as “search everywhere optimization.” I think there’s the SXO—search experience optimization—which has existed a little bit before AI Overviews, and I quite like that concept, that you optimize the actual interaction with search and not just your own result. That’s unfortunately been kind of renamed later a little bit as parasite SEO, which is somewhat similar but different as well. But I think getting that more strategic look at where we are, what our users find, how they find out about us—that might change. But the core skills: crawling, understanding indexing—may it be in any of the platforms—making sure the bots see what they’re supposed to, optimizing the information they find, making sure the information is factual but also engaging, but also does talk about our users’ pain points—that’s just SEO done really well. I feel like those who want to rename it never got that “done well” concept and were just trying to spam out their way, and now it’s not as easy as it used to be, so they are freaking out. That’s my take on it.

Peter Mead:
Okay, so they’re still not working on the things that matter, they’re still not working on the important things. I just find this really interesting, though, Sally, because maybe just to transition, talking now about—well, whether you call it GEO, whatever acronym you come up with, doesn’t really matter. At the end of the day, we’re still trying to get results from search, right? And the search can be anywhere. So it can be Google, or even if you want to go so far as to say TikTok or being found on LinkedIn or wherever, right? I mean, you still want to be—of course, we know that whole—it used to be such a great top-of-the-funnel Google, oh man, the dream, the dream of just building that top-of-funnel traffic on Google and then you turn around and say to your clients, “Look at all this traffic we’re getting to your site.” That dream, it’s really sort of unrecognized now. But I’ll actually put this to you, because recently, when I was in Sydney at the Sydney SEO Collective, we were talking, and it’s a very—I won’t mention names—but a very well-known Australian services-based website has actually been getting much more traffic from, let me say, ChatGPT and AI Overviews, because it’s all very services-based and they’re very well-structured pages. They’ve done a lot of work on the trust, on the EAT, on the entities. So instead of a lot of people saying, “Oh, we’re seeing alligator jaws,” you know, we’re seeing impressions going up and clicks going down, and this sort of alligator effect of Search Console giving you more impressions but less clicks—when it perhaps comes to being focused heavily on services rather than that sort of real top of the funnel, actually the results—maybe Google’s promise is true that you’ll get better quality traffic. What do you—would you have any thoughts around that?

Sally Mills:
Yeah, obviously I feel like it’s really tough for all sort of blog post, informational searches at the moment. I think just about anyone with blog posts on their website that previously used to get heaps of traffic, you’ve seen that whole alligator graph in Search Console, you’ve seen the impact of AI just providing the information. I do think there’s going to have to be a line, hopefully, there, because even today I was googling savings.com.au and then, say, LVR—like, looking for a specific topic within the home loan space to find a specific one of my articles. So it’s a complete navigational search, right? And then Google’s given me an AI Overview with one or two savings links and then some competitor links. So I just think it’s at this weird turning point at the moment where we’re probably going to see Google tweak things, because if the idea of AI mode becomes the whole new Google, I don’t think that’s very efficient for navigational-type services searches, where people are just trying to get to a specific—they already know about a website, they already know about a specific web page. I feel like Google’s going to end up with a bit of a blend, right? Informational content, it’s a bit sad, I don’t know how that’s going to go long term, because Google also needs us to feed the beast. So if they start stealing all our content all the time, pay overviews, people aren’t clicking through as much, we’re not seeing that sort of ROI from doing it—why are we going to continue investing in it? In which case, they’re going to have nothing to really be trained on, so it’s going to be a bit of a weird cycle, I think, and things are just going to change again and again, really.

Peter Mead:
Of course, there’s a whole bunch of new tactics. There’s quite a lot of advanced stuff—I’ve been really all theory, fan-out, all kinds of things like that. Perhaps a few things that I’ve kind of thought about is how perhaps the LLMs, how they will flatten the content to text before, you know, that kind of thing. There’s a lot of commentary around all this. I’m not sure—I got a question here about embeddings. David Carrasco, about embeddings. Maybe, Natalia, do you want to talk a little bit about embeddings?

Natalia Witczyk:
I feel like there’s a lot of smart people using smart methods that require a lot of technical understanding of how things work, and there are this other side of the curve which just copy, like, embeddings, but don’t really understand what goes behind it—just use them because it’s the hype. I think it’s really interesting how we see both sides, and people actually follow—and I really trust that they know the technical bits—saying, “Careful, just don’t take it and approach it like that,” because if you do quality content anyway, do you need that, really? Do you have to overengineer every part of every process? I’m on that side, because I don’t feel like my technical knowledge of how AI systems really work behind the scenes and how they predict, you know, where and stuff like that—I don’t think I understand enough to do it really well, and I don’t want to be on the side of the cave that just copies the overengineered topics. But I have to admit, I am doing a lot of strategy, international scaling, and maybe less detailed content analysis in my day-to-day, so maybe I don’t need to know that. But I think it’s too much of a hype for now. That’s my take on it.

Fabrizio Ballarini:
Me too, me too. Exactly. We produce our content so that our customers who read this content—who at the moment are human—understand it, like it, and engage with it. I think everything else in between, at some point, if you reduce a sentence with five words, you can do as much as you want, but it’s that sentence with those five words, right? I don’t know. I don’t think we have changed, or at least—and also, again, similar point—there could be a situation where throwing in a lot of resources might make you understand how to do something a little bit different, but also at this point, personally, I feel it’s a bit premature. It’s fine to do it on a website with 10 pages to prove something, but for us, we don’t really care. We have a lot of traffic, we continue to have a lot of content, and in the same reason why maybe in the past also, even in the case that completely we would get away traffic—and I’ve been asked this, right? My team, sometimes we get all together, everyone that does SEO, and one of the questions a while ago was, “What happens if all the traffic goes away?” Not just a little bit. And I’m like, well, originally, we are not a company—you know, like, in order to do money transfer, currency exchange, spending abroad with our card, you will need to come to Wise at some point, right? It’s not that you can do it on the model. So I think we offer a service that is quite specific, and we’re not adding a small website that explains how Wise works. Eventually, you need to come to Wise, so from that side, they’re okay.

From the aspect of informational content—we have a blog, we produce a lot of content for our customers, and obviously we teach them, we try to educate them on things that are not really, you know, money transfer. So we explain to our customers how to relocate to Singapore from London. For sure, we are not the owner of this type of service, we are not the owner of that information. Back in the days, this was adding traffic. What if that traffic would go away? Then distribution changes of that traffic. I think that’s how we’ve seen that. Even before AI Overviews, we started to diversify quite a bit on reusing the same content across platforms, especially when you go to Asia. We have a team in Singapore that does WeChat, Alipay. Forget about people clicking out of WeChat. The content is in WeChat. Actually, even the company almost stays in WeChat. You don’t buy Gucci shoes on the Gucci website, you buy Gucci shoes on WeChat. So I think there are situations where, since a while, we have been diversifying distribution. And then imagine that completely, distribution goes away. For someone that would have to start to make that content fresh, maybe as you were saying earlier, maybe there’s not too much incentive, right? Because why should I spend money to create content that then will not get distribution and I have no customers as a company? Imagine I start a company, a website today. So from that point of view, it’s true. But at least from our point of view, we have 12–16 million active customers. There must be someone of those that need to have this information that we can send it to them and they would really benefit from that, so that they can do what they’re trying to do in their life and then as well use Wise. So I think for sure there are situations where that mechanism might break, because people were—you know, every medical provider was writing medical content, like, “Hey, my shoulder is a bit painful, what could it be?” Probably that information is table stakes, right? Because I don’t think those types of websites—and I’ve seen a few getting it quite heavily—it’s not that they were really owning the knowledge around why a shoulder is painful. They basically built a massive glossary, they basically built a massive explainer. Now, probably some of this information might not be worth traffic anymore from that perspective. So I think from that point of view, it’s a bit, not sad, but at least there’s a change going on, possibly.

Peter Mead:
Okay, okay, okay. You answered a lot of questions, a lot of things that I’ve been thinking about, especially around the value of content creation and the content distribution and the targeting of that content. You answered a lot of that, Fabrizio. But what I’m wondering is, just before we wrap up before the end of the hour, I want to go around and ask each person—I’ll ask you while you’re here—how do you future-proof your approach? How do you keep everybody working and keep your whole strategy moving at the same time we’re not really sure what’s going to happen in the future? What would be your approach to that?

Fabrizio Ballarini:
Yeah, maybe I will start and then pass the ball. Definitely, compared to maybe three years ago, where I would have been 100% chilled on the fact that we do the same good old thing, we do more of it and we do it with more people, with more—I think definitely there was some level of risk on some stuff. So on future-proofing, a little bit, we will have to see maybe in the next 6–12 months what the real shift is. The fact that this technology is here—I don’t think we can go back on it, right? It’s here and we have to embrace whatever is coming. So I think the only thing that we’re doing is to try to understand as much as possible and invest as much as possible on that, even though at the same time we have to pay the bills, right? Traffic from ChatGPT in our attribution model doesn’t pay the bill of my team at the moment, right? So we have to be clear on what we have to invest at the moment, but then also, maybe if before we were not worried at all or we were not investing at all in learning something new, now we’re distributing some resources to that in a more firm way. Six months ago, we didn’t have a team just dedicated to building LLM agents. Now we have a team for that—not, you know, still a small percentage of the company, but we do.

Peter Mead:
Okay, so it’s adapting and it’s keeping up with the changes, but you’re very, very still focused very much on the whole purpose of the business, which I find—so, Sally, can I ask you, do you have some ideas, do you have thoughts about how to future-proof? Is there anything that you would perhaps say to people to think about, be prepared for?

Sally Mills:
I think the biggest thing is us just going back and trying the different AIs again and again and again, because I think they’re all changing so quickly. Even a year ago, I was very “eh,” the content it’s outputting is a bit—and just the progress we’re seeing so quickly just means that we need to go back and then try again and try again and try again. Okay, it might not have done it this time, it might not have done it to the standard we want this time, so we’ve got to do it again. So if we’re doing that, we’re not going to lose this race, I suppose, of everyone getting to the AI ship as quickly as possible. And then the second thing is just going to be human QA again and again. There’s just always going to have to be humans at a lot of steps, and I think as humanity, we need to really make sure that there’s just that human touch there, because just the risk of misinformation on the internet is—I mean, it’s already horrible, right? So it just needs to be really clear that we need humans involved still, otherwise things are going to get worse on the internet. So yeah, it’s the future.

Peter Mead:
I ask you, Natalia, how would you—would you say something to people? What can they be prepared for, what kind of steps can they take for the future?

Natalia Witczyk:
I’m going to speak purely from the perspective of someone who’s a consultant, right? So I’m external to the company I work for, and I think I already touched on it a little bit in my presentation, that human connection and meeting those people and meeting more senior people now. I think events, in-person trainings, network groups will be huge. I think this is the future, because we’re going to be so fatigued with all the AI, we’re going to need more of what we already have been seeing in the growth—like all the events that happened this year, I think it’s a record number of SEO conferences everywhere, all the time. I myself will speak at a really big SEO conference in China—Shenzhen SEO—which, you know, people fly in from all the continents to be there. So I feel like, yeah, the in-person part, so what people can do: interpersonal skills, public speaking, and gathering hands-on experience from anything you do, so then you have actual human things to talk about. My move to China is not a coincidence—I want to see how the marketing in the apps, what Fabrizio mentioned, the WeChat experience, how it looks like, because that’s our future in a way. I mean, China has been ahead of trends since forever, and I mean a couple of decades, okay, but in the digital era, since forever, and they’re going to stay ahead. So I think, yeah, gathering those in-person experiences and becoming better at interacting with other humans, that’s the future.

Peter Mead:
Wow, that’s very insightful. I just want to, you know, maybe—we’re at the top of the hour. I think we’ve covered a lot of things. Thanks, everybody, for tuning in. Thanks for the comments and some of the questions. So first of all, I want to say thank you, Sally Mills, for joining, for your opinions—it’s been terrific.

Sally Mills:
Thanks for having me, it’s been great.

Peter Mead:
Thank you, Natalia Witczyk, thank you so much for your expert opinions and your advice.

Natalia Witczyk:
Thank you.

Peter Mead:
And Fabrizio Ballarini, thank you so much—your wisdom and your wise approach, thank you so much.

Fabrizio Ballarini:
Thank you.

Peter Mead:
Okay, and to everybody, we’ll talk to you next time. Bye-bye.