AI Overview: what it is and how it changes Google search
It’s not just a new feature, but the official entry of generative algorithms into the SERP of Google, the tool that promises (or threatens?) to change the way users get answers and reorganize the dynamics of visibility for sites and content. AI Overview has officially arrived in Italy and Europe, bringing with it a whole load of doubts, concerns and curiosity about the consequences it may have for the entire digital ecosystem. One thing is certain, it will change the SERP again: what was once a set of links ordered according to semantic relevance has long since become a hybrid structure, where resources are analyzed, synthesized and presented without the user necessarily having to explore each individual source. With AI Overview there is another step forward, because Google no longer limits itself to listing results, but directly reformulates a contextualized answer, combining fragments of text extrapolated from multiple pages to offer a structured summary of the searched topic. The logic behind this innovation is clear: to reduce the number of steps needed to obtain information, by assigning the AI the task of filtering, connecting and summarizing relevant content. A transformation that speeds up searches for users, but that requires new thinking from content creators and SEOs, because the rules for emerging in the SERP are changing once again.
What is AI Overview
AI Overview is a Google Search feature based on generative artificial intelligence that provides structured answers directly in the SERP, processing information from multiple sources to offer the user a complete and immediate summary. Introduced in 2024 and enhanced by the Gemini 2.0 model, this technology allows Google to synthesize relevant content, selected from the web through a mechanism called “query fan-out”, which performs parallel searches on multiple subtopics to generate a coherent overview.
In practical terms, when AI Overview is activated on a search, the engine processes an articulated textual response accompanied by links to in-depth sources. The result is not a simple repetition of links, but a summary constructed by combining indexed documents, real-time data and structured content, with supporting graphics and multimedia elements.
With this innovation, Google’s goal is to respond to the acceleration imposed by competitors such as SearchGPT and update its search process, reducing the number of steps needed to obtain answers and providing a more fluid experience in searches that require complex summaries.
The evolution of Search: why AI is transforming the way we find information
The integration of generative AI in Search is not an isolated addition, but a process that has been ongoing for years. Google has progressively added to the traditional list of results a series of tools aimed at returning increasingly immediate and precise answers, such as featured snippets, knowledge panels or interactive carousels. With AI Overview, this trend comes to fruition, transforming the SERP into an environment where the user receives direct and contextualized summaries from the AI instead of having to build their own search path through different links.
Driving this evolution is the growing need to simplify complex content, reducing the time needed to find information without sacrificing the quality of the answers. The fan-out query system, for example, allows the algorithm of Google to fragment a single search into several subtopics, examining a variety of reliable sources to return a complete view of the requested topic.
It is clear that this transformation has a direct impact on visibility models and web traffic, because user behavior changes hand in hand with the evolution of the technology that supports them: If until a few years ago the search engine represented a point of access to information, now it itself becomes a content processor, with an increasing ability to reorganize digital knowledge autonomously – with all the consequent implications, also in terms of transparency and copyright.
The difference between AI Overview, snippets and traditional results
AI Overview represents a leap forward compared to the features already present in the SERP, and in particular they seem to be an improved and enhanced version of featured snippets. Whereas these featured snippets extract a direct passage from a single site and show it in a privileged position, the answers of AI Overview are the result of a broader processing, which combines information from multiple pages and reworks them with natural writing, also occupying a much larger space on the screen.
The system doesn’t just faithfully reproduce fragments of text, but integrates and modifies them, creating a sort of reasoned overview on the topic of the query. This technology draws from different sources to construct a coherent narrative, aggregating concepts into a unified discourse that can be enriched with images, graphics and references to structured databases such as the Knowledge Graph.
The impact of this innovation is reflected in the dynamics of organic traffic: with more complete answers directly in the SERP, the number of clicks on individual results can change, and with it the optimization strategies necessary to make content stand out within these new automatic summaries.
Availability and global rollout: where it is already active and how it is arriving in the rest of the world
AI Overview was initially introduced in the United States in May 2024, and then gradually expanded to other markets based on Google’s tests. Today it is available in multiple languages and regions, including Europe, the Middle East and Asia-Pacific, with activation planned only for users over the age of 18 and with a progressive rollout logic that adapts to the data collected over time.
Not all queries generate AI Overview and its frequency of appearance varies according to the type of search and the quality of the available sources. This means that users in different countries may experience different conditions, because the algorithm dynamically adjusts the function to ensure that the answers are reliable and useful.
How AI Overview works and what information it displays
Google’s new answers are the result of an advanced system that combines generative artificial intelligence and traditional search algorithms to produce articulate texts based on a model of multi-source analysis and synthesis. Instead of limiting itself to classifying and ordering web pages, Google now directly elaborates the contents, fragmenting them into key concepts and reconstructing them in a fluid and contextualized text.
This process takes place through Gemini 2.0, Google’s proprietary LLM (Large Language Model) that identifies relevant data, combines it in a readable and coherent format and returns it to the user already organized. What makes such an elaborate level of synthesis possible is the query fan-out approach, which breaks down a single search into several sub-queries, analyzing them simultaneously to reconstruct a response that takes into account different information perspectives. The selection of sources follows criteria of quality and relevance that are intertwined with the traditional ranking system.
Generative AI in action: how Gemini 2.0 processes and synthesizes answers
Gemini 2.0 was developed specifically to manage complex research and is able to understand and reorganize information, not just reproduce it: the model analyzes the user’s requests and determines which data can contribute to an effective answer.
Through a process of hierarchical evaluation, the algorithm identifies key elements, reorganizes the texts into a logical structure and optimizes the language formulation to offer a fluid and informative synthesis. This system differs from the extraction methods used in the past because it reprocesses existing texts to improve their readability and informational value.
Gemini 2.0 is designed to integrate information from multiple reliable sources, cross-referencing indexed content with other structured inputs, such as data from Google Knowledge Graph and data updated in real time. In this way, the visualization of the answers can include textual elements, images and even comparative diagrams, depending on the type of search.
Query fan-out: why Google divides a search into subtopics
One of the most advanced features of AI Overview is the ability to split a main query into several sub-queries, in the query fan-out process that allows Google to analyze each semantic component of the search, processing separately the information needed to build a richer and more detailed answer.
If the user enters a broad question such as “How does metabolism work and what factors influence it?”, the system does not search for a single page that contains the entire answer, but breaks down the question and activates parallel searches on subtopics, such as “Biological processes of metabolism”, “Metabolism and genetics” , “Influence of diet on metabolism” collecting the most relevant data for each aspect.
Processing through fan-out queries ensures that the AI Overview covers multiple angles of the same research, making the information more articulate than a simple text extract. The result is a more structured and in-depth overview, able to respond more comprehensively than a list of links.
Activation criteria and excluded searches
As mentioned, not all queries activate AI Overview, and the system decides when to show it based on evaluations of usefulness and reliability, separated into three main criteria.
- Degree of query complexity – AI summaries are mainly activated for complex questions that require detailed explanations or cross-referencing of multiple sources. Simpler searches, such as “Tomorrow’s weather in Milan” or “Price of an iPhone 15” , are unlikely to generate an Overview because the data is already available in structured sources.
- Confidence in the results – Google uses a reliability threshold: the AI Overview is only shown if the system has enough coherent data from reliable sources to formulate a solid answer. On topics where information is controversial or there is no clear consensus, the Overview may not appear at all.
- Excluded categories and sensitive queries – Google has determined that certain types of searches will not trigger AI Overview for security and liability reasons. These include advanced medical topics, legal issues, personal finance, and critical real-time events, i.e. classic YMYL content, to avoid the risk of disseminating inaccurate or distorted information.
The activation of the function is dynamic, so content that does not generate an AI Overview today may start to do so in the future, based on the evolution of the algorithm and the refinement of the selection criteria.
How sources are selected: the relationship with classic indexing
One of the main questions concerns the method by which Google determines which pages to include in AI Overviews. The process does not exclusively follow the classic ranking, but the sources used are selected based on thematic relevance, reliability and ability to respond to the search intent stated in the query.
Factors affecting selection include:
- Completeness and clarity of exposition – The AI Overview favors content that deals with the subject in an exhaustive way, avoiding sources with fragmentary information.
- Structured data and semantic markup – Sites that use schema markup and clear information structures can be more easily processed by AI and included in the generated overviews.
- Traffic share and authority in the sector – Pages that already demonstrate they attract qualified traffic on a specific topic are more likely to be taken into consideration. This is one of the insights gained by Ivano Di Biasi, confirmed by the first “field” analyses.
Although traditional ranking is still a determining factor, AI Overview goes beyond simple position in the SERP, favoring resources that can offer a detailed, well-constructed and reliable answer. This leads to a change in the approach to SEO: it is no longer enough to aim for first place, but it becomes essential to optimize content so that it is evaluated by AI as the best reference sources.
Initial data on the impact of AI Overview and user behavior
The introduction of responses based on artificial intelligence is already changing the relationship between users and search engines. The first studies carried out in the United States highlight a measurable change in browsing behavior, with direct implications on organic traffic for many websites: in particular, there is a reported reduction in the number of clicks on traditional results and greater interaction with the synthesized answers from Google, especially for searches that do not require detailed information.
However, the impact of the new search experience varies according to the type of query and the category of content involved: sites that provide generic information resources or quick answers seem to be the most exposed to the risk of a reduction in traffic, while other sectors react in a more nuanced way.
Fewer clicks on traditional results? What studies say about the decline in organic CTR
The initial data suggests a clear trend: when an AI-based response is displayed at the top of the SERP, the number of clicks on the links below it decreases. Recent studies on the click-through rate of informational searches indicate that pages positioned among the first traditional results record a significant decline in clicks, with more marked variations in some categories of queries.
According to research published in the first quarter of 2025, pages at the top of organic search saw an average drop in CTR of 7-10% in informational searches, while for some generic questions the effect was even more pronounced. However, an interesting trend emerges when comparing different types of queries: for those that require a higher level of detail, users continue to interact with the sites, using the previews provided as a starting point for further research.
These numbers confirm that the content synthesized by AI intercepts a group of users who would previously have explored the sites available in the search results. However, not all categories are affected in the same way.
How does information navigation change? The effect of immediate response on searchers’ habits
Direct access to a ready-to-use summary is also changing the way people interact with search engines. When a user first obtains a detailed summary, they don’t always feel the need to continue, especially when the information provided clearly responds to their needs, a bit like what already happens with the zero-click search effect.
This dynamic reduces the exploration of pages in organic results, marginalizing many resources that were previously consulted even for a superficial search. If in the past people manually selected a result based on the title and description shown in the SERP, today they tend to rely on what is reported in the summary generated by the AI.
This change doesn’t only affect the number of clicks, but also the time spent on the pages. Preliminary analysis suggests that, in searches where the AI response appears more often, the average duration of sessions on linked sites tends to decrease, indicating a less in-depth consultation. However, in some contexts this reduction in the time spent translates into an increase in the quality of the visits: those who click do so with a more targeted intent, avoiding accessing sites that do not really respond to their needs.
The sectors most affected: data from information searches and e-commerce
The emergence of AI responses has not impacted all categories in the same way: the greatest variations are seen in information queries, where users are more inclined to stop at the summary shown in the SERP without exploring further. This effect particularly affects searches related to definitions, general knowledge, general culture questions and short practical guides.
In the e-commerce sector, on the other hand, the impact varies according to the type of product and the phase of the purchasing process the user is in. When the AI responses include product comparisons, summary sheets and purchase advice, traffic to the sites that traditionally provided this information may be reduced. On the other hand, for complex transactions or purchases that require more in-depth evaluation, users continue to visit official sites before making a final decision.
The situation is more stable for queries that involve technical specializations, detailed analyses or advanced knowledge. When the topic requires complex explanations or an authoritative point of view, people are more willing to go beyond the initial answer and consider more in-depth sources. This scenario suggests that, for content producers, focusing on exhaustive and well-documented analyses could be a determining factor in maintaining visibility.
SEO and AI Overview: the new rules for ranking successfully
It’s inevitable that this functionality will lead to yet another re-proposal of the question “Is SEO dead?”, or more simply push us to reflect on how to redefine optimization strategies, shifting the focus from positioning in organic results to the ability to become the source selected by AI to generate the answers provided directly in the SERP.
The objective is no longer simply to reach the top positions in classic results, but to understand how content is chosen, synthesized and re-proposed by artificial intelligence. This means moving beyond a logic based on keywords and moving towards optimization that improves readability, relevance and content structure for AI models.
SEOZoom had already anticipated this transformation with specific analyses and targeted tools for SEO for AI, and the data collected confirms that those who structure content in a more articulated way, responding in depth to the needs of users, are much more likely to be included in AI responses, while those who rely on superficial optimization based only on targeted keywords risk being marginalized.
From keywords to search intent: why global relevance is more important than ever
Traditionally, SEO has been based on identifying the most relevant keywords and optimizing them within web pages. This approach has already shown its limitations, and is now being further surpassed by a system that analyzes queries in a more complex way: AI doesn’t just identify a direct textual match, but interprets the context of the entire search, constructing a response that integrates multiple facets of the requested topic.
AI-driven search engines interpret the complete information need of the user, and do not focus on individual keywords – and this is evident in the way that AI Overview breaks down a query into subtopics and constructs a comprehensive response by combining information from multiple sources.
To be selected, content must demonstrate that it covers the full spectrum of information on a topic, offering connections between the main and related aspects and ensuring a broader level of coverage than that required by traditional optimization. Pages that cover a topic in a fragmented way or focus on a single keyword are less likely to be considered in the AI response process. Structuring articles according to a model based on search intent, adopting a topic-first approach instead of a keyword-first approach, becomes an essential strategy for maintaining relevance, as does adopting a conversational logic and including multiple perspectives within a single piece of content.
Analyses conducted with SEOZoom have already highlighted this trend: content that effectively covers the entire range of searches related to a topic, organizing coherent and interconnected information, obtains better results both in traditional ranking and in AI selection.
The analysis of frequently asked questions, for example, is an effective strategy to align with this new model. SEOZoom has integrated tools such as Question Explorer, which allows you to identify which questions the AI considers to be priorities on a given topic. Creating content that answers these questions in a structured and coherent way improves inclusion in summaries.
Authority and reliability of content: what rewards Google’s AI in the selection of sources
The quality criteria that Google uses to select content in the AI Overview increasingly coincide with the EEAT (Experience, Expertise, Authoritativeness, Trustworthiness) framework. Artificial intelligence needs highly reliable sources to generate credible answers, which is why it rewards sites with a solid reputation in their sector.
Studies conducted by SEOZoom and the analyses in SEO for AI confirm that Google favors content that demonstrates real experience and deep expertise in the relevant sector. To increase the likelihood of being selected by the AI, it is essential to:
- Clearly demonstrate the experience of the author with structured insights, references to concrete data and contributions that go beyond the information already known.
- Adopt a writing style that is clear and optimized for AI reading, eliminating redundancies and unnecessary textual complications that can make it difficult to extract the truly relevant parts.
- Structure the content with clear formatting, avoiding excessively long and dispersive texts in favor of paragraphs organized in a logical and coherent way.
Optimize for AI Overview and don’t disappear in the SERP: practical strategies to remain relevant
To maintain visibility in a SERP increasingly dominated by artificial intelligence, it is necessary to adopt an adaptive SEO approach that takes into account the new dynamics of AI selection. From tests conducted with SEOZoom, some fundamental practices emerge to increase the probability of being chosen in AI overviews:
- Extended coverage of the topic. Accurately map the entire information ecosystem related to a topic, avoiding treating it in a fragmented way on several unconnected pages.
- Expand the analysis of the search intent. Don’t limit yourself to direct questions or dry keywords, but delve deeper into the context, anticipating the sub-topics that could enrich the discussion.
- Logical structuring of content. Use clear subtitles, well-organized lists, and segmentation of information that facilitates automatic extraction.
- Organize content with a scalable structure. Texts should encourage vertical reading, responding to multiple levels of detail. Consistent headings, logical linking of information and clarity of exposition increase the probability of being integrated into AI responses.
- Conversational optimization. Write content that answers real user questions in an in-depth, clear and natural way.
- Manage semantic optimization and data markup. Correct use of text formatting, structured data and the relationships between information elements improves interpretability by language models.
It may seem trivial, but the evolution of artificial intelligence is not only a challenge, but also an opportunity for those who can adapt quickly and create truly useful and well-structured content. The future of SEO will go beyond the fight for the top position in organic results, to become a battle to gain the trust of AI as an authoritative source from which to draw answers.
First-hand experience: what we discovered about AI Overview
Analyzing how AI Overview works based solely on Google’s official statements is not enough to really understand how content is chosen. We need to verify in the field which sites are selected, according to which criteria, and what patterns emerge in the generated responses.
This is what we have been doing for a few weeks now, as also explained by Ivano Di Biasi in this TikTok video, with a series of experiments on a wide range of queries, monitoring the behavior of the AI by observing which sites are included in the responses, which type of content is most valued and which signals seem to affect the choice of sources. Thanks to SEOZoom’s tools, we can already draw concrete indications on the most effective optimization strategies to maintain visibility in the new AI-driven SERPs.
@seozoom AI Overview arrives in Italy: how does Google change with AI? #ai #artificialintelligence #aioverview ♬ original sound – SEOZoom
The aim of these initial analyses of ranking trends was to identify recurrences and precise patterns in the results generated by AI Overview, verifying, for example, whether
- the sites shown in the AI responses coincided with those best positioned in the traditional SERP.
- differences emerged in the selection of content with respect to featured snippets and knowledge panels.
- Which factors seem to have the greatest impact on inclusion in AI overviews.
Basically, an initial significant fact emerges: the AI Overview does not always draw from the sites that occupy the top organic positions. At first glance, Google seems in many cases to prefer sources that cover the topic in its entirety, even if they are positioned lower in traditional rankings. But this is only a partial view, because the traffic share analysis with SEOZoom reveals that it is precisely the pages that win traffic on the entire cluster of keywords of the search intent that “end up” in the AI selections, and not those that are better positioned for individual keywords.
In simpler terms, the sites most frequently selected by the AI are those that, in their sector, effectively cover the entire spectrum of related searches, dominating the traffic share of a specific topic. The functionality doesn’t extract a result based solely on the position of the page in a specific SERP, but evaluates which sites are more authoritative and complete with respect to the search intent expressed by the user.
This means that, to improve the probability of being included in the AI overviews, it is essential to cover an entire semantic area, creating content that responds in a broad and detailed way to the possible variations of the same question.
The analysis also shows that Google uses multiple sources to construct answers, avoiding relying exclusively on a single page. Sites that cover a topic in depth and rank for multiple related searches are more likely to have their content extracted, while less comprehensive pages are excluded. This dynamic suggests that an approach based on topic clusters and search intent not only continues to be effective for traditional SEO, but becomes even more relevant in the context of AI-driven search.
A second relevant aspect concerns the structure of the selected content: priority is given to texts that cover a topic in an articulate way, organizing the information in clear and logical subsections. Sites that simply offer a short and direct answer, without delving into the links between the various facets of the topic, are less likely to be included.
What to expect in the future: is AI Overview just the beginning?
The integration of artificial intelligence in research doesn’t stop at these automatically generated summaries, and Google is already experimenting with new ways to make the search engine even more interactive, transforming it from a consultation tool into a system capable of dialoguing with the user and interpreting their needs with a flexibility never seen before.
The first test phases of Google AI Mode suggest that the relationship between users and results is changing quickly. The search engine is evolving towards conversational interaction, reducing the need to browse through multiple sources to find the required information. At the same time, artificial intelligence is beginning to understand and interpret not only texts, but also visual content, making images and videos increasingly central to searches.
These changes raise important questions for content creators. If AI is able to synthesize and reformulate information taken from different sources, what will be the added value of original content? The quality of writing, the ability to offer a unique perspective and the construction of experiences that are truly useful for users become decisive factors for remaining relevant in a web where information is increasingly conveyed by artificial intelligence.
Google AI Mode and increasingly conversational search: a new paradigm on the horizon?
If AI Overview represents an evolution of search, AI Mode is a real paradigm shift. This experiment, initially available only in Search Labs, introduces a more interactive experience, allowing the user to explore a topic in real time through a sequence of questions and answers directly with artificial intelligence.
The mechanism behind this innovation uses advanced language processing models to understand complex searches and handle complex, multi-level queries. The user asks an initial question, the AI provides feedback based on available sources and, if necessary, the conversation can continue with targeted follow-up requests.
This approach brings the search engine closer to AI-based chatbots, but with one fundamental difference: integration with the ranking system and the ability to include fresh, verified results, avoiding the critical issues typical of closed models that do not draw on the most recent data. The degree of reliability and the use of sources therefore remain at the center of the search experience, but the way in which users interact is becoming progressively more fluid, eliminating in many cases the need to manually consult multiple websites.
Integration between AI and visual search: why images and videos are increasingly important
The evolution of search doesn’t only concern text, because AI is improving the ability to interpret and propose visual results, expanding traditional image search and bringing new tools such as Lens and Circle to Search to the center of the online experience.
Artificial intelligence can identify patterns and correlations in images and videos, using them to provide more dynamic and contextualized answers. This approach is particularly useful in sectors such as e-commerce, education and technical assistance, where the ability to recognize an object or situation from a photo or video frame can significantly improve the accuracy of results.
Furthermore, Google’s increasing focus on multimedia content optimized for AI suggests that visibility strategies will have to adapt. Descriptive videos, images accompanied by accurate metadata, and visual resources designed to be easily interpreted by AI will play an increasingly central role in the construction of digital content.
The role of human experience: creating content that remains relevant in the age of AI
If Google is able to generate structured summaries and construct answers from selected sources, what margin of action is there for those who produce original content? The answer lies in the ability to differentiate oneself through a more authentic approach oriented towards direct experience.
AI can analyze large volumes of information, but it is not capable of replacing intuition, opinion and creativity. Users continue to look for insights that go beyond a simple summary of generic concepts, favoring content that offers distinctive value, that comes from real experiences, case studies, original analysis, technical insights and personalized interpretations .
Artificial intelligence can organize and distribute knowledge, but trust in the sources remains a key factor. Creating content that shows competence, authority and critical capacity therefore becomes the winning strategy for maintaining a direct relationship with the public and preserving relevance in a context dominated by AI responses.
