Gianluca says: “It may sound like a long shot in this exact moment of SEO history, but I really recommend that you start studying/investigating what agentic search really is.”

What is agentic search?

“It's something that, for just the last few months, big guys like Google, OpenAI, Microsoft, and other entities in the AI search space, are promoting and pushing – and it's going further than the classic use of AI that we already experienced with AI overview and SearchGPT.

It’s really about connecting an AI platform with a website through API in order to allow a user to conduct conversational search and then carry out actions directly on the website through this AI platform.

Let me give you a couple of examples. Let's say that we want to buy a flight ticket for going to the next brightonSEO, and we start a conversational search with SearchGPT or Gemini and we ask, ‘Please, find me the best flight offers for London Gatwick on this date.’ What Gemini or SearchGPT will do is visit all the websites they have in their index (as we can call it) and, through API, consultant them and retrieve the flight offers to show to us.

Then, we take a second step on the AI platform, and we say, ‘Okay, please reserve for me the flight on the Ryanair website with this offer.’ What the agent will do is go to the Ryanair website and do the process for you.

Obviously, this is a very, very superficial and simple example and things are going to become more complex. Just think about how this will work for e-commerce, for subscriptions, and for informational things. What is the impact going to be? In a sense, we are not going as a user to a website; everything is generated by AI on its platform.

My second example is related to a new application of Gemini by Google that was announced in December, which is called Deep Research. Deep Research is essentially an agent based on Gemini that allows us to ask it something like, ‘Okay, I need to do a search, an investigation, to prepare a paper about the Phoenician presence in Spain.’ What Deep Research does is consult all the indexes, understand the question, then retrieve from the index all the sources that may be important for us, present them, and tell us what those sources are.

It can create a brief – not a simple synthetic brief, but a long-form brief – about this topic for you to start creating your work. What you can do is improve the services that are presented to you by asking it to refine the sources, remove some of them, or add new ones. Then Deep Research prepares the work for you, essentially.

If you are lazy, you can use what it prepares directly for your own purpose but, if you are more intelligent, you will use it as a guide for preparing your work. Agentic search is a way to save all the time spent searching for a reliable source because these are going to be offered directly to you by the agent.

It's making things simpler for web users. This is the purpose of agentic AI and, by definition, also for agentic search.”

Is this suitable for any type of business or is there a certain business model that this is more suitable for than others?

“I think that Google and OpenAI consider that this is going to be for any kind of business. Let's look at another example related to what we could call a small local business.

Agentic search is substantially making life easier for things like, ‘I want to go to a Pakistani restaurant on this day at this hour. Tell me which are the best-reviewed restaurants in this area. Reserve a table for two people in the one you consider to be the best – or present me the restaurants, and then, after you have presented the restaurants, I can decide which one I want to reserve.’ Everything is done like this.

This is a practical example of how this agentic search can also be good for small local businesses. It's not just about a specific niche or a specific size of business. It's really something that is going to shape how search is going to be – not now, maybe not in the next 6 months, but surely in the incoming future.

My suggestion is to please start studying it and absorb all the information you can because, if you are still thinking of SEO as just a classic search environment, we're already seeing how it's been agonising for a few years now. We are in the middle of a transition passage from one way to do search into another way.”

It sounds like you're almost competing with search engines here.

How do you know where to draw the line with your business model and what you're trying to do to attract visitors to your website, without trying to compete too much with search engines that are trying to do the same thing?

“I understand your question, but I think it's starting from an incorrect perspective. A business owner doesn't really care about traffic. Traffic is a byproduct because the more traffic, the bigger the percentage of potential conversions can be.

Now, what really matters to him is potential conversions. If the conversions are coming from an agentic search or classic search, it doesn't matter. If you really work for agentic search in the correct way (at least as I'm understanding it right now), many things are the same things that we are already doing. We don't need to think in different revolutionary ways.

The fact that you may be receiving less traffic on your own website doesn't mean you are going to have fewer conversions, but more of the conversions are going to come from the agents guiding the user, on their platform, to do these kinds of things that are converting them into our customers. It's not really that different, in the ultimate outcome.”

It sounds like SEOs have to take more of a role in directing the future business model of a website. Is that a reasonable suggestion?

“Yes. This is also the real way that SEO can still be important in the future, in my opinion. Exactly as you are saying, we must direct the evolution of our approach to web marketing and organic search marketing in this way. It's transforming. We also need to reassure the clients that less traffic on the website is maybe not going to be so dangerous, if you are really optimizing for agentic search, for instance.

As I was saying, many of the things that we are already doing to rank well on classic Google search right now – like serving the correct entities, correct keyword research, and creating authoritative, trustworthy content that makes our brand stand out from the average and acquires authority, popularity, and trust. These things are going to be the same. The technical side is going to be the same because these agents are part of a search environment.

The website will still need to have technical perfection – state-of-the-art perfection – because, if not, we know that they are not going to be – let’s not say ‘indexed’ because we are not really talking about indexing in the same way. The content will not be parsed. If it's not parsed, it cannot be proposed by an agent.

Many things are going to be the same. We are going to add a few more things, like all of the work around creating APIs that can talk with the agent. This is something that, in the Software as a Service niche, for instance, is already considered – usually client to client and not thinking of search engines as if they were a client. However, it's the same thing, essentially.

Many of these things that we should consider for agentic search are going to be the same. It's really more a change of perspective than radically changing the daily work.”

If an SEO is in regular conversation with business directors about what they're doing and, traditionally, their tasks have been technical improvements to the website, bringing more traffic into the website, and optimizing different pages for organic search, and then they suddenly start talking about these newer topics to their managers, what kind of language do they need to use to justify focusing on these new areas? How did you measure the impact of these new activities?

“I’ll start from the last part of your question: how to measure the impact. This is the weak part, at least right now, because the impact in terms of traffic is not really measurable – or, if it is measurable, it's really hard to measure.

For instance, Google Analytics 4 presents traffic from SearchGPT, Perplexity, or other LLM-based AI platforms as a referral, not as organic. Search Console is not even giving us the opportunity to understand whether some of our organic traffic is coming from AI overviews, People Also Ask, or a classic search result.

Bing is telling us something about the traffic generated by Copilot but, at this moment, it's really complicated to track this. This can be a problem because, if it is not trackable, we cannot demonstrate to the client the weight that it is going to have.

I think that analytics platforms, and maybe Search Console, Bing Webmaster Tools, SearchGPT, and Perplexity, will start to have something similar, for business reasons. Maybe this year we are going to start seeing improvement in this sense. I want to think positively.

In terms of explaining it to stakeholders or managers, what I would do is not really go into jargon. This is good practice in any case. Avoid the jargon because, as soon we start using SEO jargon with stakeholders and managers, they disconnect their brains and stop listening to you. I would present the situation. I would really try to use an allegorical and metaphorical example, trying to translate things to something that they have a better understanding of.

For instance, when I was giving the example of a restaurant, it was a small business. I would say, ‘Okay, what you know is what you are doing with your clients in your shop and when you're directly talking with your clients on the phone or by email. The agent is essentially one of these clients coming to you – he's in charge of a potential client who wants to do something with your website, your products, the tables in your restaurant, or whatever it is.

What you have to do is give it all the information you would be giving to your customer when he's in the shop or calling you on the phone asking you to reserve a table. It's just a different step. We have to create the technical basis for creating the opportunity to have this conversation in the best way. Then, all the other things that we are already doing still matter.’

As I was saying: creating informational content that really stands out, giving you authority and popularity, making you a leading voice about a topic, having your PDP product pages with all the information clear so that it can be parsed correctly by search engines as we're already doing for Merchant Center.

It's going to be the same. Merchant Center is a very, very proto-version of a potential agent, essentially, because it's a layer between you and the customers, and the conversion can happen directly in the checkout of your website without even passing through the product page of your website.

We can use things that are similar to what agentic search will be in its full version, and what we already see that are ultimately similar in various executions, even if they are not really agentic.”

If an SEO is struggling for time, what should they stop doing right now so they can spend more time doing what you suggest in 2025?

“I have a pet peeve, which is keyword research as it is usually done. I don't want to say that keyword research is not important – it's really important – but I don't really understand why you have to retrieve millions of keywords to understand what the topics are and, from them, then start creating a content strategy.

I think this is really time-consuming, even if we have really wonderful tools that can help us cluster things. I think this is an upside-down methodology. If I want to do correct keyword research and spend less time, I will start with: This is the topic. What is the ontology related to the topic? Think of the solar system as a domain, for example. In this ‘solar system’ domain, we have entities: ‘sun’, ‘Earth’, ‘moon’, ‘Jupiter’, and so on.

Start by doing the ontology research to really understand what the topic is and then go down through entity search and, from entity search, start the more limited but more effective keyword research. Don’t start with millions of pins on your apartment floor and select them one by one to create groups. Start from smaller groups and then go into the really important keywords.

This doesn't mean that you have to stop chasing the long tail keywords that really matter but, if you start with an ontology and then go to an entity, then go to keywords, you are going to create content which is more likely to rank and be visible for long tail keywords that no tool is able to show you.

This is what I would say: stop wasting time on very long and time-consuming keyword research. Do it the other way. It’s faster, and it’s better.”

Gianluca Fiorelli is an International and Strategic SEO Consultant, and you can find them over at IloveSEO.net.