Google, AI Mode, and SEO: what really changes (and what doesn’t)
Confusion, unease, agitation, fear of being left behind. After Google I/O 2025, this is the mood in the digital world. AI Mode, Gemini overviews, and new forms of response within SERPs have sparked concrete doubts among those who fear a collapse in traffic, those who (still) talk about the end of SEO, and those who wonder whether it still makes sense to write articles or invest in content. The perception is that of an irreversible turning point, but the numbers tell a different story: the majority of searches are still carried out in the traditional way, and content—if relevant and structured—continues to be the starting point for the engine, even in AI responses. This is not the time to panic: it is precisely in this phase of transition that data, context, and a clear vision are needed. Because while it is true that Google has embarked on a profound transformation of its ecosystem, no, we are not facing a clean slate, and if we study how visibility is changing, we can (and must) also understand how to act to maintain it.
At the root of the fears: everything that is changing
For several months now, the mood in our industry has changed and sentiment has become negative. It wasn’t objective data that shook the ground—no abnormal traffic drops, no structural changes to how search works—but a widespread perception of a loss of reference. In many cases, the feeling is that Google has changed the rules of the game without warning, and that artificial intelligence is no longer a support, but a replacement force.
Google I/O 2025 was the catalyst that brought to the surface a tension that had been brewing for months, between the advance of artificial intelligence and the role of people in the production of digital value.
For many, it was not a conference like any other. It seemed like a formal act in which Google took a clear stance: AI is no longer a technical support, but a new operating center, a new interface of knowledge. The problem is that this narrative was intense, ambitious, but perhaps lacking in context. And this allowed fear, suspicion, and confusion to creep in.
Why is the unease so deep, if the real numbers say that the majority of searches continue to go through traditional SERPs? Because it’s not just about SEO, nor is it just about visibility: it’s about identity, role, recognition. For years, we have been building content for an engine that (at least in theory) rewarded structure, relevance, and authority. Today, an AI summary can answer a query without showing any of the original content. This creates confusion, but also anger and fear. However, now is the time to stop and take a closer look: not everything has been replaced, not everything has already been decided. Change is real, but it can be understood and, therefore, managed.
A reaction that goes beyond SEO
What we are seeing is an emotional transformation rather than a technical one. Artificial intelligence, in the collective perception, has always been a potential antagonist, but now it is emerging as a concrete systemic threat to all those who produce value through content.
In forums, on social media, and at conferences, there is less talk of strategy and more of survival. And not just among those who write for Google. The feeling of being replaceable has reached those who create, those who teach, those who analyze, and those who design. It is something deeper that affects everyone.
According to The Economist, Google searches related to “unemployment caused by AI” have reached an all-time high. In many cities, there is open discussion about how long it will be before artificial intelligence takes our place. And even if the actual data—as we will see—paints a less dramatic picture, the public narrative has already done damage: it has fueled an idea of the surpassing of humanity, which is also reflected in the world of digital content.
When fear took shape
The “problem” with Google I/O 2025 is not what was announced, but how it was communicated. Gemini is everywhere: in email, on the phone, in the browser, in search. Demos showed voice assistants observing the world and acting on it, models creating videos from scratch, systems planning for us. Gemini was presented as a ubiquitous, cross-cutting presence, capable of acting, responding, creating content, and generating media. A system designed to resolve user intentions before they are even clearly expressed. With little “scruples” about the origin of information, the fate of websites, or the potential impact on the web outside of Google.
It is this opacity about a structural change that has triggered an emotional response: fear has taken over.
What we are experiencing is a time of transition, but also of fragility. Artificial intelligence is slowly but visibly changing the way we access and consume information. Even the search experience is transforming, and with it, the ways in which we see things: AI Overview is already present in SERPs and generates responses without requiring a click; AI Mode, in contexts where it is active, further deepens this transformation.
It is unclear why content is selected for the overview, or how it is interpreted. Traditional analysis tools are no longer sufficient. And Google — so far — has not provided transparent alternatives.
At the same time, there is a growing feeling that we can no longer rely on what worked yesterday: keywords matter less, structure seems irrelevant, authority is not enough. In reality, none of these elements have disappeared. Transformation does not (necessarily) mean exclusion. Human content is still at the heart of the system today: it feeds AI responses, guides models, and structures summaries. The problem is that this link has become less visible, less immediate to read. And what we don’t see, we often fear.
Yet the direction is clear: if we know what to look for and how to interpret it, we can regain our bearings. And begin not only to understand what is changing, but also where we can intervene.
How Google is changing: what AI Overview and AI Mode are, and why they mark a change of direction
Taking the evolution that began a few years ago to the extreme, Google no longer limits itself to returning results: it is beginning to compose autonomous, summarized, conversational responses. This does not mean that the classic system has been abandoned, but that a new cognitive interface is emerging, which reworks information and offers a pre-composed version based on two pillars: AI Overview and AI Mode.
The difference is simple but decisive. AI Overview is now visible in Google’s SERPs, including in Italy. It is the block generated by artificial intelligence that appears above the organic results for certain queries, with a summary text accompanied by links to original sources.
AI Mode, on the other hand, is a more advanced experimental mode, active only in the United States and only for those who subscribe to Search Labs. Here, interaction with the search becomes conversational: the user dialogues with Gemini, explores answers, and receives contextual suggestions.
Both of these modes represent a turning point from the classic SERP logic and pose new challenges for those involved in visibility and content.
AI Overview: the generative response already active in SERP
AI Overview is the most visible, immediate, and in some ways most unsettling new feature for those working in digital. It is a block of text automatically generated by Gemini that summarizes the information deemed most relevant to a given query. It appears at the top of the SERP, before the organic results, and in some cases completely intercepts the user’s search intent, reducing the need to click on a link.
In Italy, the Overview is already active for a growing number of informational queries, even non-complex ones. There is no need to activate anything: the average user sees it, often without even noticing. It is constructed by taking content from multiple sources, which is processed by Gemini and reformulated into a cohesive text. Sometimes the links cited appear at the bottom, but the connection to the source is not always clear.
The scope of this change is profound: the answer comes before the content, and the content is no longer presented as something to be explored, but as raw material from which the AI has drawn a conclusion.
Those working in SEO cannot ignore this dynamic: we are no longer (only) competing for a ranking, but to be absorbed and cited in the Overview. And there are still no comprehensive public tools to know when, where, and how this happens—although, as we will see, something is starting to change.
AI Mode: the experimental vision of conversational search
If AI Overview is a visible effect in SERPs, AI Mode is a paradigm shift in interaction. It is only active in the United States, for those who have access to Search Labs, and represents one of the most advanced experiments in the new Google ecosystem.
In this mode, the user does not search, they converse. They enter a query and receive concise answers, but they can also delve deeper, follow up, and ask for a reformulation. Gemini acts as an assistant, drawing on the classic engine but filtering everything through a conversational interface.
The clearest difference is not only in appearance, but in intention. AI Mode aims to become the new face of search, an environment where users obtain information, perform actions, plan, and explore — without ever returning to a list of links.
It is in this environment that Google is testing its future: integration between search and tasks, personalized, adaptive, context-based responses. It is not yet a global standard, but it is the incubation platform for what we may find ourselves using in a few months’ time.
What’s really new (and how it works behind the scenes)
What makes AI Mode a game changer is not just that it responds, but how it constructs that response. The technical flow that powers the conversation is closer to traditional search than you might think, but its management is completely different.
When a user enters a question in AI Mode, Gemini is the first to respond: it analyzes the query, interprets its meaning, breaks it down into sub-intents, and triggers a semantic search. But unlike a pure AI engine, Gemini does not generate the answer from scratch: it queries the classic Google index, selects relevant content, and uses it to compose a coherent, summarizing answer in natural language.
It is therefore a hybrid between retrieval and generation: classic search provides the data, AI reworks it and presents it in a new form. The main difference is in the interface: the user no longer sees a list of results, but receives a organized output directly. At that point, they can follow up, ask for alternatives, or delve deeper.
This is precisely where Google is still experimenting. In the first versions of AI Mode, the interface was extremely linear: only text, no graphics. Today, however, there are integrated maps, product sheets, carousels, video previews, and most likely advertising blocks will soon appear as well. This is a partial but clear return to the components of the old SERP.
This tells us two things:
- Google is seeking a balance between AI synthesis and the sustainability of its economic model, which is based on visibility and interaction.
- The user experience is not yet defined: AI Mode is not the end point, but a platform under construction, which could change many times before becoming a standard.
Two sides of the same transformation
AI Overview and AI Mode are part of a single evolutionary design: to transform the search experience from a list of results to a proactive response system.
The Overview, already active in many SERPs, shows how Google is delegating the construction of the response to Gemini. AI Mode, still in the testing phase, takes this approach further by introducing dialogue, personalization, and interaction.
One shows what has already changed. The other anticipates what will change.
For those working in content and SEO, the message is clear: the model is shifting from “the right page to click” to “the right answer to generate.” And in this new context, content quality remains crucial — but it must be machine-readable, integrable into a summary, and positionable even without a visible URL. We are no longer in the era of optimization for links and rankings alone, but in the era of readability for models, clear semantic structure, and linguistic relevance.
What to do now: strategy, not panic
The phase we are going through is complex, but not uncontrollable. Google has not erased the past; it is building a bridge to a new order in which artificial intelligence becomes an integral part of the search experience.
And as always, it will not be the fastest who win, but the most lucid: those who know how to read change without passively undergoing it, adapt their strategy, and guard the right points.
A new mindset is needed. It is no longer enough to position yourself: you need to immerse yourself in the models, be readable by generative artificial intelligence, visible on multiple fronts, and consistent across all formats. We are in a phase of concrete and measurable transition, in which AI models are changing the way Google displays answers, but they have not erased the logic of content, pertinence, and relevance.
We can no longer think in terms of static rankings, but rather in terms of access: access to the generative block, access to integrated social formats, access to user attention through every touchpoint that Google considers relevant. We need a strategy that combines technique and meaning, form and authority, structure and recognizability.
SEO is not dead today. It has become broader, more interconnected, more semantic. And it continues to reward those who work methodically. And visibility does not just mean “appearing.” It means being chosen — even by an AI model — and repositioning yourself in an ecosystem that is no longer made up of blue links, but of overviews, citations, social media, and brands.
Reading and measuring the new visibility
One of the most common frustrations among those who work on SEO is the feeling of not understanding if and when content ends up in an AI Overview. For months, Google left this area completely opaque. No one knew — nor could they know — if their texts were being used to generate those summary responses.
Now something is happening. Google has announced that impressions generated by AI Overviews are starting to appear in Search Console.
It’s not yet complete tracking, but it’s the first concrete sign of openness towards more transparent measurement. Knowing that a page contributes to the AI response (even if it doesn’t receive a click) changes the perspective: visibility is no longer measured only by direct traffic, but also by the ability to be selected and integrated.
This means that we can no longer be satisfied with monitoring rankings and sessions alone. We need to look at what Google considers useful at a semantic level, even when it doesn’t generate traffic. And work accordingly.
Writing for artificial intelligence
Writing “for AI” is not a new literary genre. It’s a change in mindset: we need to produce content that is easily readable, interpretable, and synthesizable by a language model.
What does this mean in practice? It means:
- Structure texts with logical clarity: one answer for each question, cohesive paragraphs, meaningful titles.
- Use natural, but precise language: no convoluted sentences, no ambiguity.
- Pay attention to grammar, punctuation, and formatting: AI rewards formally correct texts.
- Avoid noise: remove redundancies, fillers, and rhetorical excesses that confuse models.
The goal is not just to “rank,” but to be considered useful enough to be reused. This is truer today than ever before: content is read—by artificial intelligence—before it is read by a user.
Integrate positioning: website, social media, brand
This change began long before Google I/O 25 and even before AI Overview. Those who follow SEOZoom know this: we have been saying for years that positioning needs to be rethought as an ecosystem, no longer as a list of results. Not (only) because of intuition or privileged information, but because we have always analyzed every signal, read the variations, and listened to the field.
And the transformation was already visible in the data, in query behavior, in SERP patterns: increasingly, the content that Google highlights comes from mixed environments: social posts, short videos, product pages, public profiles, visual carousels. It is the answer to a simple need: to offer the best information available, regardless of where it resides.
This changes the rules of the game. Or rather, it expands the playing field. Those who limit themselves to working only on their website risk being left out. Those who also maintain social media channels, produce videos, and write in a consistent and recognizable way have more points of access to visibility. And if there is one element that Google tends to favor in all its AI interfaces, it is brand recognition. All other things being equal, the winner is the one who is known, quoted, and visible on multiple fronts.
SEO today is no longer just “Search Engine Optimization.” It is Search Everywhere Optimization, as Ivano said at SEOZoom Day 2025, it is “visibility architecture,” or building a solid, credible, and interconnected presence that speaks well to both users and models, everywhere.
What changes (and what doesn’t) for content
Artificial intelligence has changed the way Google displays information, but it has not eliminated the centrality of content. Let’s repeat that.
AI Overview did not come out of nowhere: it is built from texts published on the web, analyzed, selected, and synthesized by Gemini. The source of truth—albeit increasingly less visible—remains human content.
It is true that the model decides what to take and what to ignore. It is true that in many cases the answer is consumed without a click. But it is equally true that only well-written, clear, and complete content is used to feed these answers. In other words, AI does not eliminate editorial work—it becomes a new reader. And if the reader changes, so does the way we write to be chosen.
The task now is not to chase formats or technicalities, but to rethink content as a semantic unit that can be read by a model: orderly, useful, structured, easily summarizable. Those who know how to do this still have a place. Those who continue to produce for the sake of filling space risk becoming invisible.
Content does not disappear: it fuels the response
The most common mistake is to believe that AI Overview has ‘replaced’ content. In reality, it absorbs, processes and summarises it. Gemini reads pages, identifies salient information and constructs a coherent summary. If content appears in that output, it is because it has been judged useful, credible and relevant.
That’s why textual quality is no longer just an editorial issue, but a strategic lever for visibility. It’s not enough to “be on topic”; you have to treat it comprehensively, coherently, and in an up-to-date manner. Overviews feed on content that truly answers questions, explains key concepts, and doesn’t force the reader (and the model) to look elsewhere.
This is not the time to write less. It is time to write better.
Grammar and clarity beat keyword stuffing
“Forced” techniques no longer work. Texts built around repeated keywords, bloated sentences, and useless lists are penalized upstream because Gemini better recognizes readable, well-formatted content that is free of redundancy.
We are not just talking about stylistic elegance: we are talking about semantic accessibility. A linguistic model processes grammatically correct texts better, with well-distributed punctuation and a natural logical rhythm. If a concept is clear to a human reader, it will also be clear to AI. And the opposite is also true.
Writing for visibility today means also thinking about the automatic interpretability of the text. Order beats cunning. Simplicity beats cleverness.
Keywords remain relevant, but their role is changing
Some have been too quick to declare the “death of keywords.” This is a misguided exaggeration. Keywords remain the gateway through which an AI system understands the intent of a query and structures its response. But their role has changed: from labels to semantic signals.
It is no longer enough to repeat a keyword to be relevant. It needs to be placed in a coherent context, where every word reinforces the meaning, not simulates it. Gemini is not fooled by forced density or fake titles: it recognizes credible linguistic patterns and selects what really has informational value.
Those who continue to work on keywords as if they were a filter to be fooled are preparing for a search engine that no longer exists. Those who treat them as semantic keys, using them to build texts that truly respond, explain, and connect, will still be visible. Even tomorrow.
Where visibility ends today
The question is no longer just “how do you get on the first page,” but where does visibility end today. Because Google no longer returns just web pages: it returns videos, social content, images, cards, carousels, stories. The SERP has become a mosaic. And those who want to be there must think of themselves no longer just as owners of a website, but as active nodes in a network of presences that Google considers authoritative.
It’s not about “doing social media,” but about considering every channel as a potential entry point into search. Google has already integrated YouTube videos, TikTok clips, Instagram posts, and Pinterest content into its results. And it doesn’t put them at the bottom of the page: it brings them to the top, connects them to informational or commercial intent, and treats them as answers.
Being present in these contexts is no longer optional. It is a concrete part of organic visibility.
Google integrates social media, video, and visual formats
Just do a product search, a local query, or even a simple informational question to see how much Google’s results have changed. Instead of ten blue links, we find YouTube videos, image carousels, TikTok and Instagram posts, visual cards, and embedded content. The engine no longer thinks in terms of pages, but in terms of formats: it searches for content where it knows the user spends time, attention, and trust.
This transformation is now stable: visual and social content are no longer exceptions or experiments. They are structural parts of SERP, with their own detection and positioning logic. For some search segments, the “social” part represents over 40% of total potential traffic.
Positioning yourself today means being visible on multiple channels
For years, “doing SEO” meant working on your website. Today, that’s no longer enough. Content producers need to think multichannel: Google also indexes what we publish on other spaces. And if that content is relevant, useful, and well-structured, it can rank even better than a web page.
This does not mean wasting energy. It means designing content with cross-platform reuse in mind from the outset. A guide can become a script, an excerpt, a reel, or a visual. A well-expressed piece of advice can live on as a tweet, a short video, a slide, or a podcast. And in all these formats, Google can intercept it and present it to the right user at the right time.
Positioning today is a distributed system. And SEO becomes visibility architecture, not just page optimization.
SEO for social media: from hypothesis to reality
It used to be said that “social media cannot be positioned.” Today, the opposite is true: social content is positioned in SERPs, often before traditional editorial content.
But for this to happen, it is not enough to publish. You need to write, shoot, and title with search logic.
A script designed to be read by an algorithm must:
- contain the right keywords, used naturally;
- be structured with a clear introduction and explicit promises;
- visually show what it says, in a way that is consistent with the user’s query.
This applies to a video on YouTube, a reel on Instagram, or a slide on LinkedIn. Anyone who produces social content today without a search engine logic is only speaking to half their audience. Those who manage to get Google to read them there too are expanding their visibility — with the same effort.
Work and AI: why the apocalypse can wait
Fears about artificial intelligence do not stop at SEO or content creation. In many areas — from consulting to customer care, journalism to logistics — there is a widespread belief that AI is replacing human roles one by one.
Yet, the data does not support this scenario. The impact of AI on work is still limited, progressive, and more related to the transformation of tasks than to the elimination of jobs.
Global research on “unemployment caused by AI” has exploded, but behind the anxiety there is often a communication short circuit: perception runs faster than reality. Those who stop at the headlines see a world where workers become superfluous. Those who look at the numbers find a different story: work is changing, but it is not disappearing. And for many, it is becoming more powerful.
The data contradicts the anxiety
Today’s labor market does not reflect a massive substitution effect caused by AI. In the United States, unemployment is stable at around 4.2%. Wage growth continues. In OECD countries, the employment rate is at an all-time high. And even in sectors theoretically “at risk,” such as translation or data analysis, the number of people employed has increased compared to the previous year.
The same is true for clerical jobs, often cited as the most vulnerable to automation: according to data reported by The Economist, the share of employment in those roles has grown. In other words, AI is not emptying offices, newsrooms, or agencies. It is entering workflows—but as a support, not a replacement.
AI adoption remains partial (and cautious)
Another important finding concerns the actual adoption of AI in companies.
According to official estimates, less than 10% of US companies use AI to produce core goods or services. Many have started experimenting with it, but few have integrated it into their core processes.
This is because truly integrating AI requires time, culture, resources, and profound organizational change. In most cases, AI has been introduced to automate support activities, speed up internal processes, and generate drafts or prototypes—not to replace entire roles.
The reality is much more nuanced than the mainstream narrative: companies are investing in AI, but they are not laying off workers en masse. They prefer to optimize. And often, they do so together with people, not against them.
AI changes work, it does not eliminate it
Of course, there is an impact. Some activities will change. Some roles will be redefined. But talking about “the end of work” is out of scale today. AI is a tool: powerful, yes, but not autonomous. It does not make decisions, it does not handle complex ambiguities, it does not communicate with empathy.
What it can do is increase productivity, help generate preliminary content, summarize, and suggest alternatives. But human value—in selection, strategy, and relationships—remains central.
Those who work in content, marketing, training, or consulting should not fear the disappearance of their role. Instead, they must update their skills, integrate new tools, and position themselves as interpreters of change, not victims.
AI changes the ways we do things. It does not erase the ends.
Opinions are divided: some cry collapse, others observe
In the debate that followed Google I/O 2025, the reactions were not all the same. Some saw Gemini and the AI Overview as the beginning of a total disintermediation of the web. Others took a more cautious approach, emphasizing the experimental nature of many features and the importance of distinguishing between announcements and real impact.
This is not a clash between optimists and doomsayers. It is a comparison between different levels of interpretation, all legitimate, responding to different sensibilities and priorities.
On the one hand, there are experts who read the signs with strategic concern, highlighting the consequences that the transformation of SERPs may have on publishers, independent sites, and content creators. On the other, there are analysts, journalists, and technicians who seek to bring the facts down to a concrete scale, avoiding excessive projections.
It is not a question of deciding who is right. It is about understanding what elements are really needed to read the present and prepare for the future.
Giorgio Taverniti’s analysis
Giorgio Taverniti is one of the most influential voices in the Italian SEO scene, and he dedicated the entire Fastletter n. 55 (May 26, 2025) to post-Google I/O, which was widely discussed. He did not mince words: for the first time, he explicitly referred to Google as “standing as a publisher,” and doing so “on too large a scale” to be ignored.
His reasoning starts from a very clear point: Google is no longer just a search engine, but an entity that organizes, synthesizes, and redistributes content. The introduction of AI Overview—and even more so AI Mode—leads to a new information structure where the original source is no longer the center of the experience, but a secondary, optional reference.
Answers are generated in-house, using material published by others. The links are there, but they no longer guide navigation. They are attachments, not gateways.
This is not just a technical issue, but an economic and strategic one. If the answer is consumed in Google, visibility is concentrated there. And with it, economic value. According to Taverniti, the risk does not only concern publishers or SEOs, but the entire information ecosystem: from independent blogs to e-commerce, from comparison sites to agencies.
Added to this is the prediction—already partly confirmed—of a future in which even advertising will become part of AI responses. At that point, Google will:
- Take content from the web
- Summarize it with Gemini
- Present it in its interface
- Earn money from advertising
- Drastically reduce the need to visit websites.
What he describes is not a malicious intention or an ethical distortion, but a structural transformation. Today, content no longer serves only to inform a reader: it also serves, and increasingly so, to feed a linguistic model. And those who create it, in this scenario, must rethink their role, their strategy, and their value chain.
This is not generic alarmism, but a lucid and concerned reading of current dynamics. Taverniti observes that visibility is not disappearing, but is concentrating in places that are no longer controllable by those who produce the content. And that Google, even when citing sources, does not always guarantee a return in terms of traffic, authority, or conversion. That is why we need to work to be integrable, citable, and visible in environments that do not display content, but digest it for the user.
The point of view of those who downplay
A different, more analytical and measured position comes from Alex Kantrowitz, founder of Big Technology, who believes that “the disintermediation effect is real, but it is not as widespread as people think.” Google has shown a vision, not an operational reality, and AI Mode is still a confined experiment, not a default change. Kantrowitz points out that much of the attention has been captured by isolated demonstrations, not concrete usage data. He urges us to distinguish between strategic vision and immediate impact to avoid disproportionate interpretations.
Rand Fishkin, founder of SparkToro, also highlighted another critical point: “Google used to cite 7-8 sources for an answer. Today, one is enough. AI Overview rewards the few and cuts out the many. It’s not just a question of traffic, but of concentration of information power.” It’s a clear view that doesn’t deny the transformation but makes us wonder who will survive and why in this new visibility hierarchy.
On the other hand, Barry Schwartz, a long-time expert in the search world, emphasized the lack of transparency: “We still don’t know exactly why a page is selected for the overview. There are no public tools or clear guidelines. That’s the real problem.” In his post-event video, Schwartz calls on Google to offer visibility and control tools to those who produce the content that feeds Gemini. Without these guarantees, there can be no talk of balance.
Google’s voice: between narrative and operational reality
From within Google, the line is clear: Gemini is the new center of search. Already in his keynote speech, Sundar Pichai said, “We want to build an engine that understands what you need without you having to think about how to formulate it. Our goal is to reduce complexity.”
And to begin responding to growing criticism from publishers, analysts, and trade associations, Google has chosen a clear line of communication: downplay the impact of artificial intelligence on organic traffic, arguing that the web is more alive than ever. On several occasions, the company’s public figures have reiterated that Gemini—and AI Overview more generally—does not generate answers out of thin air, but is based on the reworking of content already published online. The quality and structure of the texts on the web therefore remain a central element in the selection process.
Pichai himself, in a post-I/O interview with The Verge, stated that “we are sending traffic to more sites than ever before. No one sends traffic to the web like we do.”
The CEO of Google (and Alphabet) responded directly to criticism from publishers and the media, in particular to the harsh statement from the News Media Alliance, which spoke of “content theft” and called for the intervention of the US Department of Justice (DOJ). In his “defense,” he reiterated some key points:
- AI Mode and AI Overview will continue to include sources and send traffic to websites.
- Google remains “the company that cares most about the web ecosystem,” even compared to new platforms that do not address the issue.
- Outbound traffic from Google is expanding to a wider range of sources and publishers than in the past.
- Despite the change in interface, the informational and commercial value for sites is not disappearing, but evolving.
“Five years from now, we will continue to send a lot of traffic to the web. That’s the product direction we’re really committed to,” he said, also talking about “higher quality referrals” thanks to overviews: according to Google, the time spent by users and their interaction with AI-selected content would indicate more relevant traffic, even if potentially distributed differently.
When asked why the publishing ecosystem is experiencing a period of severe economic tension, Pichai replied that change can lead to redistribution: some sources receive more, others less, but the overall amount grows. It is “the dynamics of evolution,” and in his view, “if Google didn’t provide value, users would go elsewhere.”
The interviewer replies that if everything were really going better, publishers would not be so angry. But Pichai closes without backing down: the direction has been set, and the product will continue to evolve in this way. He repeats two figures to support his argument: the increase in the average time spent on sites reached via overviews and the growth in the number of indexed pages (+45% in two years), elements which, according to him, demonstrate vitality, not compression.
Reassurances (on the surface) or a reaction to the context?
Google has every interest in presenting Gemini as an omnipresent, fluid assistant, useful at any time of day.
The demos seen at Google I/O show it looking at the world, understanding what is happening, booking flights, writing messages, and organizing the user’s life. It is a powerful narrative—and probably true: Mountain View’s AI seems to have quickly overtaken Openai and other competitors, even if in the common perception generative AI is still the same as ChatGPT—but it is also strongly marketing-oriented.
That’s why we need to look at the facts: AI Overview remains a feature that only activates on certain queries. It is not always present, it is not yet optimized, and it has not replaced the classic engine. And even AI Mode is currently limited to a small number of US users.
Google is building a conversational ecosystem. But the beating heart of search is still anchored to external content, to texts published by people on websites, which must be readable by AI, integrable, and structured.
As long as this is the case, those who produce information value will still have a role to play. And the goal today is to keep that role visible, accessible, and recognizable. Even within a changing interface.
And we must not forget to analyze the broader context in which Google operates and the other actions it is considering. Consider, for example, the strategic move that now seems ready to be unveiled: as we reported, Google Discover is also coming to desktop, bringing a stream of recommended content similar to that already visible on mobile.
This is an additional source of potential traffic, which looks very much like a possible ‘sweetener’ for publishers at a time of high tension – give with one hand, take away with the other.
This tension is not only media-related, but also political.
Google’s public defense — “traffic is growing, the ecosystem is holding up” — is very reminiscent of the company’s defense in the lawsuit filed by the DOJ in the US, which concerns Google’s control over the search market (and, in parallel, another lawsuit dealing with the ads segment).
The charges are abuse of a dominant position, restriction of competition, and concentration of traffic and revenue. And any change in the interface, traffic sources, or monetization models can also be interpreted as an attempt to balance interests and pressures.
Between AI Overview, new desktop features, and continuous public reassurances, Google is trying to reinforce the narrative of an expanding, uncompressed web. But the real impact—on websites, content, and the information economy—will only be visible in the medium term.
This is the current picture: Google is changing, but it has not yet replaced anyone. What is needed is not to choose between optimism or pessimism, but to equip ourselves with the tools to read, monitor, and react. With clarity. And method.
The numbers to add to the debate
To avoid empty talk, however, we too must frame the current situation in its context and add proportion and data. In the scenario analyses we read—especially the pessimistic ones—the most advanced demos, the most ambitious promises, and the most spectacular applications are taken and projected onto the entire user experience.
The result is that we talk about research as if it had already been completely rewritten, when in reality over 90% of global queries still go through the classic SERP.
Then there is a gap in real numbers: we do not know how many users are actually interacting with AI Mode, nor how often AI Overview completely replaces clicks. Impressions are only now beginning to be tracked, and Google has not yet provided comprehensive tools to monitor them.
In this context, the task of those working in digital is neither to minimize nor dramatize, but to maintain a proportionate approach. The transition has begun, but it is gradual. And there is still room to read, interpret, and react strategically.
Let’s bring some order: what searches really tell us
The debate has already become polarized: some are crying out about the collapse of the web and the end of SEO, while others (a minority) see only opportunities in AI Overview, AI Mode, and Gemini.
Rarely are numbers cited, but this is necessary in order to put things into perspective, contextualize, and read in proportion. Data and statistics paint a less definitive picture—but one that is no less interesting.
For example, a recent study by OneLittleWeb paints a clear picture: in March 2025, the world’s ten leading search engines averaged 155 billion visits per month, compared to just 4 billion for AI chatbots. This huge difference shows that, today, traditional search remains by far the dominant tool for accessing information.
Google alone generated 1.6 trillion visits in March alone, accounting for almost 90% of global search traffic. In second place, far behind, is Bing with 60 billion visits, up +27.7% year-on-year. On the AI side, ChatGPT (including SearchGPT) dominates with 47.7 billion monthly visits, followed by DeepSeek and Gemini, both around 1.7 billion.
In short, traditional Google Search generated over 33 times the traffic of ChatGPT in the same period. And even the much-maligned Bing—the search engine that has always been “snubbed” because it is considered irrelevant—still recorded 26% more traffic than AI.
Of course, chatbots are new and still growing exponentially—search as a whole saw a minimal decline (-0.51% annually), while chatbots grew by 80.9%—but at the moment there is still no collapse of search engines, which in fact still have overwhelming supremacy. Even with this surge, chatbots account for just 3% of total traffic related to search and information. They are not yet standard, but a rapidly expanding niche.
This gap is also due to structural factors: search engines are deeply integrated into browsers, operating systems, voice assistants, and mobile interfaces. They are familiar, immediate, and automatic.
Chatbots, on the other hand, require more conscious interaction, dedicated access, and familiarity with a new paradigm. For many users, they are still optional tools.
That’s why talking about “AI overtaking” today is incorrect. Generative artificial intelligence is growing. But its impact on everyday life is still limited. And above all, it has not replaced engines, it is forcing them to change.
Of course, we must add to Google’s traffic the 47 billion AI Overviews generated in the same period: a significant absolute figure, but also marginal in percentage terms. This means that 97% of interactions with Google in 2023 still took place in the way we have known for years: with a query, a SERP, a click.
Not only that, but AI Overviews are only activated on a subset of queries, and they do not always replace clicks. In many cases, they precede, accompany, or stimulate them. And AI Mode has not yet been launched, and we do not know what its real impact will be. Yet, many post-I/O readings have treated it as if it were the new standard face of search. This is a stretch. AI Mode is a test environment. Important, yes—but still in the observation phase.
People are looking for answers, not overviews
Another key finding emerges from semantic analysis of queries.
In recent months, AI-related searches (such as “what is Gemini,” “how does AI Overview work,” “will Google remove sites from results?”) have exploded. But they do not indicate heavy use of overviews. They indicate curiosity, fear, and a need for explanation. The average user is looking for information about AI, but is not yet moving into the AI experience.
The direction is clear, but it has not yet been taken by the majority of people.
A transformation in progress, not yet complete
All these numbers should be read for what they are: they do not deny change, but they do put the perceived urgency into perspective.
We are not facing a complete reversal, but a transition phase, made up of tests, partial rollouts, and scenarios that are still being defined.
Understanding this is essential for those who work with visibility. Because it means that:
- There is still room to adapt.
- Historical metrics still matter.
- Overviews are not the only answer.
- The average user is still finding their way.
We must not ignore AI. But neither should we rush towards a future that, for now, exists more in keynote speeches than in everyday queries.
Reading change to respond better
So, is everything okay? Can we relax and continue doing what we were doing before?
Of course not: change should not be endured, it should be understood. The transformations introduced by have not (yet?) eliminated visibility, but they have reconfigured it. The rules we knew have not disappeared: they have evolved, becoming more distributed, more semantic, more selective. We need a strategy that combines technique and meaning, form and authority, structure and recognizability.
This is not a regression, it is an evolution. Those who can adapt still have room to maneuver. But not with old habits. We need a new method: analyze, rewrite, diversify.
What to do (methodically)
The first mistake to avoid is reacting instinctively. There is no need to chase every feature or rewrite everything. We need to understand what Google considers relevant and how it values it in the new interfaces.
AI Overviews do not replace content: they summarize, organize, and select it. And Google does this by reading the right signals: clarity, structure, usefulness, consistency.
In concrete terms, summarizing what has already been written, the “new” rules for this phase are:
- Don’t stop writing, but write better: content that can be read by a language model, not just by a human reader.
- Don’t rely solely on rankings: also monitor indirect visibility, integration into overviews, and semantic presence.
- Don’t remain isolated: work on integrated presence across your website, social media, videos, and brand. Google looks for widespread authority, not just well-optimized pages.
Modern SEO is a work of visibility architecture.
The new SEOZoom: built first, designed for now
It is precisely from this vision—broad, distributed, systemic—that we imagined and built the new SEOZoom. Not as an emergency response to AI, a post-I/O patch, or a reactive move out of fear. But as a platform already designed to read visibility wherever it manifests itself.
We saw the signs before they made headlines, read the changes in the data, and followed how Google was changing. We analyzed, integrated information, and even our own insights.
Today, SEOZoom is evolving with a clear goal: to help you understand not only where you rank, but where you are truly visible. Even — and especially — when the click doesn’t come, but the exposure does. It doesn’t just tell you where you rank, it shows you where you really existfor engines, for models, for users. And it helps you be there, always.
The new SEOZoom combines ranking, overview, social, brand, and discoverability in a single integrated, semantic, operational view. It’s the tool you need today if you work methodically, if you know that visibility is no longer just about position: it’s about presence, selection, and recognition.
Everywhere is not a slogan. It is the new reality of SEO. And you need a tool that can see it all.
FAQ: our concrete answers to the most frequently asked questions
We have told you what has happened, explained what is changing, and analyzed how to react. But after every explanation, questions remain. Many. Legitimate. Recurring.
In this section, we have collected the most frequently asked questions we receive every day from professionals, content creators, brands, and entrepreneurs. Because understanding is the first step to acting with clarity. And because, today more than ever, we need less fear and more answers.
- Will AI replace SEO?
No. SEO is not disappearing: it is changing form. The goal remains the same—to intercept user demand—but the path has been enriched. Today, SEO also includes visibility in AI-generated content, presence in visual and social formats, and brand recognition. It is no longer just about “climbing the SERP,” but about being found in a distributed ecosystem, where AI engines read, select, and synthesize. Those who know how to adapt remain visible. Those who insist on using old patterns risk invisibility.
- How do you gain visibility in AI answers?
By writing content that is useful, organized, explicit, and easy to summarize. AI Overview draws from web content, but it doesn’t show everything: it chooses what it can read clearly and considers relevant. There’s no need for new technicalities. We need to get back to basics: answering a real question well, without frills, without keyword stuffing, with structure and precision.
- What role do social media play in SEO today?
A central one. Google integrates content from YouTube, TikTok, Instagram, LinkedIn, and Pinterest directly into SERPs. Any content that intercepts a search intent can become a result. And every interaction—likes, comments, quotes—strengthens the brand’s visibility in AI-based contexts. Social media are not an alternative to SEO: they have become one of its sources.
- Can I see if I appear in the overviews?
Not yet, but soon. Google will show the impression data generated by AI Overviews in Google Search Console. It’s not complete tracking, but it’s a concrete signal: it means that overviews are officially measurable, and therefore optimizable. Monitoring those impressions allows you to understand if your content is being integrated, evaluate “invisible” visibility (that without clicks, but with exposure), and make decisions based on data, not assumptions.
- How do you measure visibility outside of Google Search?
With tools that read signals across multiple channels: social, brand, citations, direct traffic, Discover, AI overviews, carousels, e-commerce. Classic organic traffic is only part of the picture. To see the rest, you need platforms that combine SEO data, social data, and semantic signals. That’s exactly what we’re building with the new SEOZoom: a unified view of visibility, designed for the web that’s coming.
