Understanding AI-Powered Search – Global Lessons from Italy | Ivano Di Biasi
Welcome back to The Search Session! I’m your host, Gianluca Fiorelli, and today I’m joined by Ivano Di Biasi—Italian SEO specialist, software developer, and CEO & co-founder of SEOZoom. Together, we explore how AI is reshaping search behavior, why visibility now extends beyond websites, and how SEOs can adapt measurement and strategy to this new, multi-surface search world. Ivano shares key discoveries from his latest research:
Search habits are shifting—AI now drives product research, cutting informational traffic while transactional queries stay strong.
Google ranks web pages and UGC differently, so SEO must span web, video, and social to stay visible across all surfaces.
Pixel-based SERP data shows exposure, but branded searches remain the clearest measure of real impact.
AI works in two modes—training data or web results—and when it cites, it’s because sites already rank in regular search, not from independent recognition. In both cases, it’s still SEO—just adapted to how AI retrieves information.
Keyword clustering defines modern search—know how Google groups intents, and let each page fit its natural cluster.
Ivano foresees a structured, citation-rich SERP ahead—where AI insights coexist with user trust and publisher value.
Discover all this and more in this insightful episode.
Video Chapters
Transcript
Gianluca Fiorelli: Hi, I’m Gianluca Fiorelli, and welcome back to The Search Session! Today, we're talking with a really great gentleman from Naples. He’s the CEO and Project Manager of SEOZoom — an Italian SEO software, or software for SEO, if you prefer.
I’m really happy you’ll get the chance to know him better — or maybe even for the first time. Because honestly, there’s something special about discovering someone new for the first time. His name is Ivano Di Biasi. How are you doing, Ivano?
Ivano Di Biasi: Hi! I’m good, thanks.
How AI Is Changing SEO in Italy
Gianluca Fiorelli: So, Ivano, let me start with the same question I ask all my guests: How’s SEO treating you lately?
Ivano Di Biasi: It’s funny, actually — we’re coming to it a bit later because it’s just been a few days since AI Mode was released here in Italy. So, it’s bad news for us.
But actually, I’ve seen that not so many people are using it, because you have to choose AI mode just for certain queries.
I have to say, the big change happened when AI Overview was released here in Italy, because we saw a lot of traffic going down — and going nowhere. The big problem is that the traffic didn’t move to other platforms or other places... it just disappeared.
So, here in Italy, we had this big problem: trying to understand what we have to do to maintain all that traffic. Actually, I’ve seen websites going from 8 million visits in a month down to 2 million. So it’s a big problem to manage.
We also have to say that the old SEO niche is moving toward AI monitoring — trying to understand if we are ranking inside LLMs and AI search engines.
But actually, I don’t believe it’s so important for us right now, because every single search engine with AI doesn’t provide so much traffic.
So it's less than Bing at the moment. So, we are just focusing on SEO — traditional SEO — but we are keeping an eye on the future. If organic traffic is declining and it’s not being replaced by LLMs, we need to understand where that traffic is going.
So maybe in the future, that traffic will come from LLMs and AI search engines. So we are moving to optimization of our content for LLMs too — because maybe it will be the future. So we are doing that.
Gianluca Fiorelli: Yes, and regarding the impact of AI Overview in Italy — and for our friends listening to us and watching us — I think we must remember that Italy has somehow always been used by Google as a testing place.
Which is easy to understand, because Italy is quite a big market for Google, and it's just one language, which is not used by anybody else.
So for instance, I remember back in the days — I think it was 2016 — Google started to test the tags in image search, and then rolled them out to everyone in the world. We were the first ones.
And I’m seeing something similar now in the percentage of how many AI Overviews are used. There was a recent research showing that Italy was the country with the highest percentage of AI Overviews in the SERPs.
And I shared also, quite a few days ago, a case — I was doing a search for a client of mine, and I ended up on a search results page with two blocks of "People Also Ask." All the first eight questions were answered by the AI Overview — not the classic snippet from a website.
Ivano Di Biasi: With the links, yes.
Gianluca Fiorelli: Which is wild. And also, that's not the only place where Google is testing AI Overview. There’s another feature called “Things to Know,” which is a really nice feature — and even there, I remember one case where I had my main topic and four subtopics, and two of those subtopics were answered by the AI Overview.
So if I was basically expanding all the AI Overviews, I had a SERP so long, I could go to sleep and use it as a blanket — that’s how big it was! And actually, you can imagine why someone may decide not to click anywhere.
Ivano Di Biasi: Yes.
Gianluca Fiorelli: It even has a page bigger than a Wikipedia page in the search results!
But I agree with you about the traffic. We always talk about Google — maybe because Google has always been the biggest source of our traffic — but very few people really criticize something like ChatGPT or Perplexity, which are sending even less traffic than Google.
One of the fantastic things — and one of the reasons for the friendship I have with you, and why I invited you to The Search Session — is that with SEOZoom, even if it's really focused on the Italian market, I think it's a good proxy for what’s happening in markets all around the world.
You have a very wide vision of what’s happening. You just said that one website went from 8 million visits per month down to 2 million.
I imagine that at SEOZoom, you’ve started analyzing different types of industries. What industry was hit the hardest? I mean, publishers were surely the most impacted — it’s their niche — but what about eCommerce? Because I’ve started to see AI Overviews, for instance, showing up even for transactional or commercial queries.
Ivano Di Biasi: Yes, the big problem is that people are not searching for keywords in the same way anymore.
Actually, when you make a search, AI is triggered when you're doing tests about products. So maybe you want to find the best products with the same characteristics. In that case, when you have to make comparisons, AI comes into the field and replies to you — doing all the work you used to do manually.
I remember in the past, everyone with an e-commerce site would build a blog to try to match all the long-tail informational keywords. Then, from the blog page, they would send users to the product page to sell.
Now, all the blogs of e-commerce sites are dropping in traffic — because AI replies directly about the product, about the comparison of products.
The big problem is that in the past, we had to search on Google for product specifications, then click 10 links, open 10 tabs, read all the articles, and try to understand which one was the best fit for our needs. So it was very expensive in terms of time. Now, AI comes in at that point — to help with comparisons and to help you find something you might really be interested in buying.
So, I believe that e-commerce sites are losing traffic with the whole approach of creating a blog and informational content. But when it comes to the final part of the funnel, people go on Google, make their search for that product, and just look for the one who sells it at the lowest price.
So, in those searches that are transactional, nothing has changed. Because people get the information from AI, and then they go through regular searches on Google. This is the big change.
But I believe there’s another big change — and it’s about the real user intent that Google is trying to understand. Because when I search for something — let’s say, I bought a new house just one year ago, so I had to buy the kitchen, the bed, and all the things I needed. And when I was looking for inspiration, I searched for the “best kitchens with island” — with the island in the middle.
And Google understood that I was not looking for who was selling the product. It understood that I was looking for inspiration, because in my search — in my query — I wasn’t saying exactly, exactly what I wanted.
So it replied to me with UGC content. It replied with galleries from Pinterest. So I had to go through the photos, get my inspiration, and then go back again to do another search.
What we actually see on the Google SERPs is a representation of what the intent could be when a user makes a search. And it takes into consideration not just UGC content, but also social media.
So, if I’m looking for something that requires a tutorial — because it’s easier for me to watch a video instead of reading a page — actually, the SERP is not just made of pages. Because the page is not the right medium to reply to my intent, my intent could be hiding the idea of watching a video — or considering opinions on social media.
Gianluca Fiorelli: Yes, and that’s also why Google came to a deal with Meta in this case — to present Instagram Reels. It’s not really indexing them — it’s like opening the firewalls of Instagram in order to show Reels from professional accounts directly in search. They show up along with Shorts in the Shorts box.
This is also an interesting thing, what you just said helps explain why — for AI Overview, but also for AI Mode — the most used third-party platform for being visible is YouTube. Because whenever Google understands that the intent is visual — visual learning — it pushes that content to the front. So, I want to learn, I want to understand something — but I want a video because…
Ivano Di Biasi: It’s just the media.
Gianluca Fiorelli: Google has YouTube, and so it takes YouTube immediately and pushes it on the AI surface, so I totally agree.
From Clicks to Visibility: What to Measure
Gianluca Fiorelli: So, if clicks are not a real outcome right now from AI Search on Google, the magic word in the last months has become visibility. So, we are still trying to be visible — to have some kind of impact, as if Google were like a billboard in the street. To be there, present, with our brand eventually cited or mentioned in the AI Overview. To stick in the memory.
So, there are several questions. First: how can you really track visibility in a way that is meaningful — and not just random data about visibility?
And second: how can you say, “OK, visibility is fine, knowing the data about it is fine, but what is the real outcome of visibility?” This is what I want to measure.
Did you find a way? Because it’s a problem — even though there are people like Duane Forrester, for instance, who recently wrote a very interesting post in his Substack about how to evaluate and calculate the value of visibility — it’s still something that nobody has really figured out.
Ivano Di Biasi: I have to say that there is a curiosity I want to tell you — and it's something that happened to me personally. Because I’m not just the Project Manager of SEOZoom, but I'm the Head Developer too. So when I create a software, I try to understand what’s happening on Google. There was one moment — a perfect moment — when Google made changes to its SERPs. So, the HTML of the SERPs.
And as you know, every SEO software that creates a Google HTML to find the results will stop working in that exact moment — because the HTML changed. So, you have to rewrite some portions of the code to try to match the new results.
And what happened was very funny — because in that moment, actually, all our SERPs in SEOZoom were just populated with YouTube and social media results. So, what’s my idea?
My idea is that Google has two different algorithms: one to rank web pages, and one to rank UGC content. I believe they are separate algorithms.
Why do I think that? Because when they changed the HTML part of the regular SERP — the web pages — we couldn’t match those results anymore. But in the same SERP, we could still match the other part — Quora, Reddit, YouTube, Pinterest, and Instagram.
So everything that was from other channels — that are not regular pages — because they didn’t make the change at the same time. That part still had the old HTML, which wasn’t updated yet. Then, when they made the full update, we could see all the SERPs again together.
So, I believe that when you make a search on Google, Google can understand what regular pages are and when it wants to inflate — inside the regular SERP — other kinds of content.
Actually, I made a study, and I understood that the other types of content — from YouTube, Instagram, Pinterest — use other algorithms to try to rank inside the regular SERP. And their positions are always between the third and the seventh positions. So they’re always in the middle — or at the beginning, or in the middle — of the SERP.
So, actually, doing a regular SERP just for web pages is very, very difficult. Because, you understand, if I try to rank in the second or third position for a SERP where Google decides — arbitrarily — to inflate a YouTube video, it’s very, very difficult to work for visibility.
So, actually — coming to visibility, as you asked me — we have to consider this: because those videos could be our videos. And they don’t just bring traffic on the platform where you publish them — but you can know in advance that videos about a topic can rank in that position for the keywords you want.
So, you can continue doing regular SEO, understanding that if you create a video or a post on social media, you could rank in the regular Google SERP and take the CTR of that position to the video. So what do we have to measure?
We have to measure our visibility — the brand, the brand positioning. So we have to do multi-channel activities — and we can do that with SEO too.
We can choose what kind of social media content to create, starting from SEO ranking inside the SERPs. So, I believe this is the best strategy to cover all the channels.
Gianluca Fiorelli: Yes, it’s what I call — stealing the definition from Rand Fishkin — doing zero-click marketing, zero-click SEO on the SERPs in order to be visible in the potential features that can be triggered by Google with your omnichannel content strategy.
In fact, this is something that can be done with Advanced Web Ranking — but I think also, I’m not totally sure, maybe you can confirm — and also with a tool, a software like SEOZoom. You can actually know how many pixels you are occupying because that is maybe the value of the visibility.
Let’s say you’re visible in a search result maybe twice: once in the AI Overview as a source with that very tiny icon — the link icon — and once in the regular search, with your search snippet.
Ivano Di Biasi: Yes.
Gianluca Fiorelli: Actually, the mention of the link in AI Overview is substantially hidden within all the information in the answer of AI Overview.
And the search snippet maybe is your ranking — I mean, for position, but for position after AI Overview, after a video box, after the first two organic search results, after discussion and forum block. So, substantially, nobody's going to come to your search snippet because it's so buried on the first page that nobody's seeing it.
Gianluca Fiorelli: But having the information about how big in pixel length the page is, and knowing that of this page, you're occupying, let's say, 55% of SERP — is good information for me.
What is tricky is how you can then measure the impact of this visibility. And I think that could be branded search, maybe, or direct search, or direct type-in.
Ivano Di Biasi: I believe both. Because if you are ranking in an AI, and your brand is mentioned every time when you are looking for some information about your topic — if your brand is mentioned — people will go to regular Google and will search for your brand.
Actually, I believe this is the only thing we can track to try to understand if our brand is getting more searched. So I believe this is the only metric we have to measure, to try to understand if the AI is giving some value to our presence inside the searches.
And we were talking about pixels — and in SEOZoom, we made the same thing. We have the pixel position of the first position, the second position, and every element in the page. So, you can try to understand how many keywords I’m ranking for — show my first result out of the first scroll. We have this metric that is out of the first scroll. You see a webpage with the results, but the first result is outside the scroll — you have to scroll to see the first result.
This is because Google publishes everything inside the pages, and they are very crowded: knowledge panels, advertising, AI Overview, FAQ, and so much other information. It’s very difficult to try to understand if it deserves to rank for that keyword, if you have to waste your time ranking for something that has a very low visibility. So, this is what we are doing at the moment about the visibility of a single keyword.
You can arrange all the keywords out of the first scroll, and you can understand, “Okay, the tracker tells me I’m getting 10,000 visits — but if half of those keywords are out of the first scroll, you’re actually making less than half.”
Gianluca Fiorelli: Yes, your CTR is surely smaller than the theoretical one.
Training Data vs. Real-Time Data in AI Search
Gianluca Fiorelli: Now, going on the other side — so moving from Google and going to its competitor right now — which is not Bing, but ChatGPT. And you recently published a video, a quite long video…
Ivano Di Biasi: Half an hour.
Gianluca Fiorelli: Yes, and I’m sorry, but it’s in Italian for everybody listening to us. But I don’t know, maybe you can try this AI dub…
Ivano Di Biasi: Yes. HeyGen — it’s with the cloned voice from Eleven Labs. You can make it in English easily, and it actually speaks better than I do here, live, in English.
Gianluca Fiorelli: I was thinking even more simply — with the YouTube features. Also, many influencers started to use that AI-generated dubbing — from Spanish to English, or vice versa.
Ivano Di Biasi: I believe that Google makes itself. I believe there is an option — that you can translate.
Gianluca Fiorelli: It's an option in YouTube Studio that you can activate.
So, going into that video, you did something that it was very clear. You explained things as they are, without going too much into jargon. And you were substantially explaining something that — I don’t know — maybe it’s considered something that everybody should know, but actually, not so many know: the difference between asking something, having a conversation on ChatGPT and ChatGPT — or any other LLM — considers that it's not necessary to retrieve fresh information. So it's just relying on the training data that it has.
And this is maybe because in some cases it’s a fact — let’s say, when Julius Caesar was stabbed in Roman history — this is a fact. It’s not going to change. There’s no need for fresh content.
But instead — when the RAG and grounding are needed — the LLM has to pull in more recent information to offer the best answer.
So just quickly — for all the friends who can’t watch your video in Italian — could you give a clear definition of those two concepts?
Ivano Di Biasi: Okay. Let’s start with your first question. You said that not so many people know what is happening when they do searches in AI — and it's true.
But I have to say that I study every day. To manage an SEO tool, you have to know everything about what you're selling to your customers.
So the first problem was to try to understand what was going on in the future. I believe it was one and a half years ago when I wrote a book. It’s SEO for AI. It was my book about my idea that, one day, search engines basically based on AI would be born. I imagined that Google would be changed in the future — and that something would start using AI.
It was one and a half years ago, and there was nothing of what we are seeing today in search engines. So I had to study a lot of things, so I could have clear ideas on how LLMs work.
Now, let’s make the first difference: if we talk about AI Overview — it uses regular search, because it’s an overview. Its task is to take all the results in a regular Google search and summarize the response to the user. Therefore, that’s not to be considered as something that is an LLM. It’s just a RAG. The RAG just makes a simple model. Let’s call it a model, but it’s not. It takes 10–12 pages and tries to make a summary of what you would have to read in all the pages. So it's its task, and we can do something to rank better in that reply.
Actually, everyone is inventing new acronyms — so for that, people are talking about AEO — Answer Engine Optimization. Why can you optimize something for that answer engine? Because if you reply exactly what the user is asking — and you rank for the main keyword people are searching — then you can also rank for the questions you are replying to inside the AI Overview.
Then there is another kind of search — that is, AI Mode. AI Mode is completely different, because you’re not just searching for one keyword. There is a reasoning model that tries to understand what you mean when you make your search.
Actually, I tried it in Italian. I searched for "torta di mele" — and that in English is “apple pie”. So I searched for apple pie, and the reasoning model understood that I was looking for something generic. So, when someone is looking for apple pie, they’re usually looking for the traditional recipe of the apple pie.
As such, it made seven, maybe ten searches — different queries — found all the pages, and then replied to me with results that were completely different from the ones ranking in the regular Google search. Websites that were in first, second, and third position were totally ignored — because the AI understood that my intent was completely different.
And that is another job to do — and I don’t know if we have to call it AEO or GEO or something like that. In my mind, it’s SEO for AI. And you do the best you can — it’s not necessary to find other acronyms to try and understand this.
Another thing is when you go on ChatGPT — because ChatGPT is not intended to be a search engine. So, the web search is just triggered for certain queries — not for all the queries, as happens in the Google ecosystem.
So when you go to ChatGPT and make a search, if it needs to make a web search, the web search is done. If the web search is not needed — because the response is already in the knowledge of the model — then it replies to you directly.
What changes is when ChatGPT makes a web search and when it doesn’t. When it doesn’t make a search, ChatGPT is like a person who studies for years and learns a lot of things, reading a lot of books — and when it gives you a reply, it doesn’t need to mention you. In which book and on which page did it find that information? It just knows it. So it replies to you with the information it has.
When it uses the web search, it mentions and cites websites. And that is the point where it makes a difference. But if ChatGPT is mentioning or citing a website, it's not an added value to your strategy. It just mentions you and cites your website because it’s in search.
So what are we talking about? GEO, or are we talking about SEO? When an LLM mentions you — it's SEO. Nothing different from what we’ve been doing for years.
Gianluca Fiorelli: Yes, you're totally right.
Ivano Di Biasi: I totally believe it — because actually, sorry — there are a lot of tools, and every day a new tool is released that tries to make your analysis about how your brand is mentioned inside the LLMs. But for me, it’s nonsense. Because if you are mentioned, you are ranking.
It’s the only case when you are cited or mentioned by an LLM: it only happens when you are ranking for one of the keywords it searches on a regular search engine. And I believe that we — as software developers — are polluting the information of the SEO. It’s a dangerous moment.
Gianluca Fiorelli: Yes, and I think that there is a — I mean, it's weird — because neither you nor I is a native English speaker, but I think that many people — even native English speakers — confuse citation and mention.
When someone is mentioning — like, I’m mentioning, for instance, the brand Ferrari — a model can mention the brand Ferrari without linking to Ferrari or to any other website talking about Ferrari. But it is mentioning Ferrari.
When it’s citing something, like saying, “I’m citing this,” as if it were between brackets.
Ivano Di Biasi: Talking about you.
Gianluca Fiorelli: Yes — with a link to the source. This is when the RAG version of ChatGPT is in action.
And when we talk about answers without citations, without links to sources — I think that’s where you can try to do something. In a sense, it’s just another facet of SEO — it’s doing digital PR, doing branding for SEO, in order to be cited by third-party websites.
You can also do all the spammy things that we’re seeing — people creating tons and tons of top 10 lists of something...
Ivano Di Biasi: Yes, I’ve seen it.
Gianluca Fiorelli: ...in thousands of websites that once were selling links are now selling this kind of top 10 list, or creating the lists themselves — where, casually, they are the first one.
So, you can try to work in that way — thinking ahead, for when this training data is going to be replaced by a new set of training data. Because now, the training data — like many years ago, when we had the famous Google Dance — because Google was indexing, and for a long period, it wasn’t indexing anymore.
So the results were just those that had been indexed — and then you do a new batch of indexing. So fresh content was coming. And somehow, we are in the same situation with this specific type of search by LLMs.
I agree with you when it comes to the RAG part — so when they are going on search — doing searches on Google, using Bing, using their API, I don’t know what ChatGPT is using at the end. It’s not so clear.
Then it’s SEO — and it’s probably more interesting, eventually, to try to understand — as you were saying, your example of apple pie — what kind of query fan-out ChatGPT, or in the case of AI Mode, Gemini for AI Mode, is doing?
Ivano Di Biasi: It is extremely stupid to try to understand how it works — because it relies on Gemini. So it works the same way — you can just ask it.
Keyword Clusters, Fan-Out & Intent
Gianluca Fiorelli: Yes, but I think — and I don't know if you agree with me — when it comes to things like query fan-out, which is also another new technique, an advanced tactic that's coming out, people are doing with query fan-out the same mistake they made with People Also Ask.
Ivano Di Biasi: No, it's not the same.
Gianluca Fiorelli: Looking literally to the question — and just trying to answer only the question. What would you really suggest to these guys still doing this?
Ivano Di Biasi: Okay, I was talking Italian. I have to say that this is a problem we studied — I believe — five years ago, when we stepped out from keyword monitoring to webpage monitoring. So, if you write one content, you don’t have to do it just for one keyword — because you can rank for hundreds of keywords.
So when I was developing the core of the software, I had an idea, and I tried to do reverse engineering of Google’s behavior. I wanted to try to understand for which keywords Google was providing a very, very similar SERP — about 60–70% of the same webpages across different keywords.
Why did I do this? I did this to try to figure out: when I choose one keyword — maybe "apple pie" — I need to know for which other keywords I can rank with that same page.
So we found SERP-similarity for hundreds and hundreds of keywords, and we discovered a cluster — a keyword cluster — to try to understand when Google replies with the same kind of pages.
And actually, we found these clusters. And that cluster, after five or six years, is exactly the fan-out. So it’s what AI will search to figure out what kind of content you have to rank in its results.
So this is the way I believe we have to do it: make reverse engineering of Google and try to understand all the keywords you can rank for a subset, for a cluster of people’s interests. So they can make it in different ways.
Someone could say, "Yes, but actually, people are not searching by keywords inside AI engines." But the reality is that when you search for something there, Google can’t have a database like the keyword one — like keyword search volume for each month — because they are infinite. So it’s not feasible. They can’t have a database like the one of keywords.
That’s why, actually, Google only relies on the keywords. Also, if the technology is changing — and you search the way you want, with mistakes, with the wrong pointings inside the text — when it comes to making the real search, Google steps down to keywords, searches them, and gives you the reply.
So actually, the most important thing is the keywords — the subset of keywords — you have to rank for, with your content, to satisfy the user needs in the AI too. So this is it.
Gianluca Fiorelli: And I would say, in order maybe not to be too literal with the term “keywords” — keywords plus the intent Google may infer by your search. Because this is even easier with conversational search for Google, surely from all its algorithms and machine learning that it has developed to understand what is the real intent behind this conversational query.
And for searching those kinds of search results pages for those specific keywords that can be retrieved by the conversational query — and according to the intent. And that’s why, when you were showing the example of apple pie before, it was showing many, many pages that were ranking in position 2, 3, 4 — that weren’t used as sources — because in the query fan-out of AI Mode, maybe it exactly understood the kind of intent.
It did very specific searches, but triggered a very different type of search results from the classic ones you would eventually see in a classic search based on single short keyword research.
Ivano Di Biasi: Yes, yes. They are completely different. And, coming to the example of the apple pie, there were just three pages mentioned. So it didn’t need to find a lot of pages — it just chose three, because the intention was to find the traditional one. So if it’s traditional, they are all the same. So I choose two or three, and it’s okay. So, it's different, the approach that Google has.
But I believe we are making it so difficult — when it's actually extremely, extremely easy. Because we create content, so, that webpage will rank for a hundred keywords. I don’t want to care about intent — because the keywords are already there.
So, if I want to rank for those keywords, I write that article. If I want to rank for other keywords, I know that the article is not the right one. We have to write another one — with a different focus — to intercept other user intents, other buyer personas. So, it’s easier than it may seem. It's not so difficult to understand.
You can analyze a competitor’s webpage, okay? You can choose one of their pages, see all the keywords it’s ranking for — and all the keywords their competitors are ranking for. So, you find a subset. And if it doesn’t match what you want to reach — because you have other interests, or you want other users who don’t match that kind of user — then, okay, it’s easy to understand: you need to write other content. Not for that keyword — because you will never rank for keywords outside of that subset.
And this subset — it wasn’t invented by me. It’s a subset of what Google is doing. It’s not my algorithm — I’m just looking at Google. And Google is ranking those pages for those keywords.
So, if I want to rank for another keyword, my content is not good. I cannot just modify it — because I can’t insert a new keyword into that subset. I have to write another article. I have to go in another direction.
The Future SERP: Web Guide-Style, AI-Enhanced
Gianluca Fiorelli: Yes, totally. And, let’s talk a bit — because you said, I think you’re also passionate about search, not only because it’s your job to always work on the core of SEOZoom — about the future.
I have two questions. The first — what do you think the future of AI in Google Search will look like, considering that what we see now is surely not the final version?
And the second question — which you can answer later — what about all this agentic search that’s starting now with n8n, and with agents, I don’t remember the entire name of that thing invented by GPT. These are two different things, but sometimes people confuse them or think they are the same.
What are your thoughts about these two — not completely different, but for me, converging topics?
Ivano Di Biasi: Okay, let’s answer the first question. What I believe is coming… and the funny part is that, when I wrote my book SEO for AI — I don’t know if I drank too much that night — but I wrote it all in one night, and I had a kind of illumination.
Gianluca Fiorelli: I remember that you finished the book on a train — from Naples to Bologna.
Ivano Di Biasi: With the phone.
Gianluca Fiorelli: When we were in a conference.
Ivano Di Biasi: Or maybe I was drunk.
Gianluca Fiorelli: Probably yes.
Ivano Di Biasi: Okay, so let’s say that I figured out that when it would happen — if it would happen — that there would be an AI engine, it would be something very, very different. Because the AI would steal content from websites, so websites would not be cited or mentioned. It’s more like the old version of Google. And it happened.
And then I figured out that it could change everything about the business model — because of ads, ads inside websites. If you click on websites, and clicks are lowering, then every website owner will lose money. So this is what is happening now.
But I also said that in this future — which is not fully here yet — Google cannot give you results the way it's doing now with AI Mode. Because it's not good — not for them, and not for publishers either, so it will change.
Actually, I believe AI Mode is just like a tool — a version that you can use if you want to “talk” with Google. Let’s say it that way. But it’s not a real product — because Google still needs to use Shopping. It needs to use advertising.
And I know that in the USA, something is happening inside the replies. So if you ask something, they are providing ads, I believe, and videos, and something inside the result. But I believe that we'll see Google going back to regular SERPs, but made like a document.
So, the SERP will be created by Google with paragraph, paragraph, paragraph — titles and descriptions — and citing every website where they made the first snippet of text.
Gianluca Fiorelli: So basically, you're reminding me of something that I talked about recently in an episode with Tom Critchlow. And we were really, really praising the test lab of Web Guide — because Web Guide is a good, balanced use of thick search. Because it filters the web and enhances it with AI. What it does substantially is transform the search results page into a pillar page.
Ivano Di Biasi: Exactly.
Gianluca Fiorelli: You can see the classic 10 blue links — because it’s the filtered web of Google Search — but then you can see the Web Guide. So, there is an introduction created by Gemini, as if the query is treated like a topic. There are sections for the subtopics that Google considers most relevant for that query.
It does a sort of query fan-out also for Web Guide. So, you have this — it’s like a pillar: introduction, then section, section, section about subtopics. And every subtopic has just a short introduction, then the search snippet.
And instead, we started to see it in the U.S. in regular search — where instead of a classic meta description, Gemini is summarizing the content of a page, giving you a reason why that result is relevant for you.
Ivano Di Biasi: Yes.
Gianluca Fiorelli: For that search, you did it. I think that whenever they figure out how to insert advertising — because it would be a contradiction — they cannot put advertisements under a filtered web, because that is supposed to be just regular web search results, with no ads, no SERP features, and so on.
So, I think they are testing all these things in AI Mode. Because actually, the answer of AI Mode — even if it’s not clearly presenting the links and search results — still has a main title, then section, section, section. So, in this sense, it’s similar.
Ivano Di Biasi: I believe that will be the way.
Gianluca Fiorelli: So, whenever they understand on AI Mode — which is a live playground for them, for everything related to AI — where to put things. They already know how to put video,and they’ve started to know how to put images. They are experimenting with some shopping features, and sometimes also with Google Maps — so local search information.
So, whenever they are able to fix all these things, what I see is Web Guide being pushed into the main search — just as a web guide. “Web” will remain like the classic web filter, without advertisement, without anything — just organic search results. And AI features, first tested in AI Mode, pushed into regular search.
Ivano Di Biasi: Yes, but we have to know that when Google released the AI Mode in the USA, it had to do that. It was not its choice. Google had no choice — it had to release AI Mode because of ChatGPT. What was the alternative? Lose every customer because they’re moving to ChatGPT, or release a poor product — because AI Mode is absolutely a poor product. So, they published something that is not complete — it’s just an interface. You make a query, and it replies. So, nothing so special.
But I believe that’s just because ChatGPT was pushing and gaining new users. So Google had to do that. Actually, it’s not what we’ll see in the future.
Agents, MCPs & Monetizing AI Mode
Gianluca Fiorelli: And what about agentic search? Which is, let’s call it Semantic Search 3.0, or 4.0 — because it’s a very big implementation of semantics with engineering and so on. I don’t know how many businesses will ever really be ready for something like this.
Ivano Di Biasi: I don’t know if you mean agentic search for internal searches.
Gianluca Fiorelli: I mean, I think that it can be both. I mean, especially for e‑commerce, and especially inside first, but I’m thinking of these custom OpenAI, ChatGPT apps.
I just read today that Google is preparing an API for something quite similar to what's happening in AI Mode — which could be interesting. In AI Mode, we already see features like checkout options, and now even instant checkout directly from search. It’s not quite a true agent yet, but it’s clear that the platforms are moving in that direction. So I was wondering — on your side, is there any development toward this? And more broadly, what do you think the future looks like when it comes to the connection between these emerging agents and the search platforms themselves?
Ivano Di Biasi: Okay, I made a few MCPs. We have the SEOZoom MCP, so you can use it inside ChatGPT and Cloud. You can make searches and ask for data from a database that’s coming from an external source. But I believe that is something really personal. I don’t believe it is something that could be inserted inside a regular search.
So, when it comes to AI Mode—if they do give the ability to connect an MCP—the user will have to choose which MCP to use for their search. I believe that's something that’s not under our control. We can’t monitor it, and we can’t really understand what’s going on in every single search, especially when it comes to personalized use of the LLM.
Now, imagine if, in AI Mode, Google decided to connect, say, a Booking.com MCP. Why not Expedia? Why not smaller websites? Why not LastMinute.com? Would they have to integrate every MCP to be fair to all publishers, or would it make no sense?
So, I believe it will be up to the user—they’ll be given the chance to decide which one to use in their searches because Google would otherwise have to integrate every MCP across every niche, which I don’t think is realistic. I think it’ll be something users can choose to use.
Gianluca Fiorelli: I’m just brainstorming out loud here, but maybe this could actually be a good way for Google to monetize. I’m offering you this idea now!
We already have in the regular SERP—for example, hotel blocks that are essentially local, paid, and customized results for hotels. Same with flights—well, we don’t have that in Europe anymore, I think, but in the US it’s still very strong.
So maybe something like this—an MCP box or opportunity inside AI Mode—could be a way to sponsor or monetize AI Mode.
Ivano Di Biasi: There's a technical difference, because if I use AI Mode, it's a way to search Google. An MCP is a way to search for something else—to search for information in other places. So, I don’t know how Google could match this inside AI Mode. I believe that's something you could do more easily in a normal LLM—maybe in Gemini—but in AI Mode, I can’t really figure out the real utility of that.
So, let’s say I connect the SEOZoom MCP or Gianluca Fiorelli MCP inside Google’s AI or AI Mode, and I make a search—what would Google do? Search Google, and then mix that information with what's coming from Gianluca Fiorelli or other sources? I don’t know how that would help. Maybe it could work just for product searches.
Gianluca Fiorelli: Yes, I was thinking about product search.
Ivano Di Biasi: Like if I want to buy a car. For shopping, maybe it's useful.
Gianluca Fiorelli: That’s why I was thinking of it as a paid feature.
Ivano Di Biasi: Yes.
Gianluca Fiorelli: As an evolution of Google Shopping, or of Google Hotels, or Google Flights. Instead of just presenting the price and the link to do the booking—for instance, on Booking.com or their website—you could do the whole process directly.
You’d choose the one you want—Booking, Expedia, Hotels.com, LastMinute, Trivago, whatever—and once you click one, it would start the interactive part of the dialogue with the agent from that service, but on the surface of Google. And obviously, Google having a percentage as it has now.
Ivano Di Biasi: Yes, but when Google makes a choice, it can cause huge damage—or give huge gifts—to someone and not to others. So, every Google choice is very, very dangerous.
Gianluca Fiorelli: I know, I know.
Ivano Di Biasi: Old newspapers in Italy don’t know how to make money now.
Gianluca Fiorelli: And I must say that also many newspapers now—maybe not the biggest ones, but many newspapers in Italy—are really looking for… I don't know if it's available in Italy, but the possibility to nudge the user to choose us as their source, your favorite source. You know, that option that exists for news sites?
Ivano Di Biasi: In Discover, you can.
Gianluca Fiorelli: Yes, but also in the Top News, in regular search, I think. That you can preselect and decide. “Okay Google, I select this —the New York Times, the Washington Post, Los Angeles Times, whatever—as my preferred source and then add whatever else you want”.
So I think these are maybe just little things Google is trying to offer to the news industry in order to, let’s say, try to be pardoned.
I mean, they also announced at one point that Google Discover was going to be available on desktop. And honestly, I still haven’t seen any Google Discover on desktop.
Ivano Di Biasi: Yes, I’ve never seen it here.
Gianluca Fiorelli: No, nobody’s seen it.
Ivano Di Biasi: Okay.
Gianluca Fiorelli: It was announced more than six months ago.
The Fireside Questionnaire
Gianluca Fiorelli: Okay, so—well, we’ve been talking for almost an hour now. Let’s just conclude by putting SEO aside for a moment and talking about you.
Ivano Di Biasi: Me?
Gianluca Fiorelli: Yes, so, one of the reasons Ivano often uses the apple pie example—and why I usually go for mini painting queries, which is my personal passion—is that he tends to use cooking and recipe-related queries. It's like what I once said to Rand Fishkin: Rand is less famous than his wife. And in a similar way, even though Ivano is very well-known and respected in Italy, he might actually be less famous than his wife, Flavia Imperatore.
Ivano Di Biasi: Yes, it’s real. It’s true. But it’s because the niche is more mainstream. I mean, cooking—everyone eats. No one does SEO. So it’s for the audience. Same for us.
Gianluca Fiorelli: So, Flavia is a very well-known food blogger in Italy. A big shout-out to Flavia if you’re listening or watching us! Now, let’s talk about your beginnings. Every SEO of our age has a very different beginning.
Ivano Di Biasi: My beginning in SEO was fun, because one day my associate told me, “Do you know there’s an SEO World Championship?” And I asked him, “What is SEO?” Then he explained to me what SEO is, and I tried to attend this championship. And at the end of the championship, I was ranked second.
So, I started by understanding that link building, at that moment, was the most important thing to do. So I created some free WordPress templates, gave them away for free, and I ranked very, very high in the SEO World Championship.
After that, I became known here in Italy—and also in other countries—because I got, as my first customer, Expedia. So I went from winning the SEO Championship to having a top-class client like Expedia. It was very funny. That’s how it all started—just joking around, playing. For me, it was a game.
Gianluca Fiorelli: Speaking of playing—you always tell me, and all our friends when we’re hanging out, that you have a past as a video game developer.
Ivano Di Biasi: Absolutely.
Gianluca Fiorelli: And eventually, if you had the time, it's something you'd like to go back to, right?
Ivano Di Biasi: Okay, I have to say—it is my hobby. But actually, it's been about 10 years now, since SEOZoom started, and I haven’t had time for a hobby.
I started when I was 16, writing video games in Assembly and C++. I made a few games, and I worked for a company that brought virtual reality to Italy. That was around 1996, 1994.
Gianluca Fiorelli: Yes.
Ivano Di Biasi: I’m old enough. I’m 52, so I’ve got a long time behind me. I started with video games, and I’d love to do it again in the future—but it’s very, very expensive. So, I believe it will stay a dream; I’ll probably never do it again. But I’d like to do it. I sell everything—I sell the company—and I make games!
Gianluca Fiorelli: Well, that could be a good idea. Maybe a potential future, who knows? And you also like to travel. What’s a place that surprised you the most? Maybe somewhere you didn’t have big expectations, but you visited it and went wow.
Ivano Di Biasi: There are so many. The one that made me say wow was the waterfalls in Argentina.
Gianluca Fiorelli: You went to Iguazú.
Ivano Di Biasi: I went to Iguazú, in a small plane—it was a stormy day, so it was terrifying. Very, very scary. But when we arrived in Iguazú, I believe it’s the most beautiful thing I’ve seen in my life. So, it was very, very beautiful. Then this summer, I went to South Africa—to the Cape of Good Hope—and it's the second most wonderful place I’ve seen in my life. It’s extremely beautiful. My wife loves to see animals, so she loves to go anywhere there are animals. And I am the most scared person in the world when it comes to animals!
Gianluca Fiorelli: I can imagine you doing a safari.
Ivano Di Biasi: Yes, I did two safari trips a day, for a week, in the morning and afternoon. With no coverage for the Jeep—it was open. So, lions and hyenas coming around; it was very, very scary! But my wife loves it, my daughter loves it—so I have to do it.
Gianluca Fiorelli: Yes, I can imagine. Thank you, Ivano. It was a real pleasure to talk with you, and I’m sure we’ve shared many interesting things—also with the audience.
Ivano Di Biasi: Thank you! And sorry for my English—but it’s also funny.
Gianluca Fiorelli: Well, your English is not that different from mine. I’m not Shakespeare, so it’s fine. It was really good.
And for you, remember to ring the bell to get notified about new episodes, and don’t forget to subscribe to the channel. Thank you, and see you next time!
Podcast Host
Gianluca Fiorelli
With almost 20 years of experience in web marketing, Gianluca Fiorelli is a Strategic and International SEO Consultant who helps businesses improve their visibility and performance on organic search. Gianluca collaborated with clients from various industries and regions, such as Glassdoor, Idealista, Rastreator.com, Outsystems, Chess.com, SIXT Ride, Vegetables by Bayer, Visit California, Gamepix, James Edition and many others.
A very active member of the SEO community, Gianluca daily shares his insights and best practices on SEO, content, Search marketing strategy and the evolution of Search on social media channels such as X, Bluesky and LinkedIn and through the blog on his website: IloveSEO.net.
stay in the loop