Google AI Mode: search is no longer the same, it is a conversation

What Google AI Mode is and when it arrived in Italy

For decades, searching on Google meant the same thing: typing a few keywords, scrolling through a list of ten blue links, choosing which one to open. A process that, while evolving with snippets, knowledge panels and AI Overview, had always kept the same underlying logic: the user searches, Google suggests sources, the user decides where to go.

Google AI Mode breaks this logic structurally. It is not an incremental update — it is a paradigm shift. Instead of showing sources, Google answers directly. Instead of a list, it generates text. Instead of a search, it starts a conversation.

Built on the Gemini model, Google AI Mode was announced at Google I/O in March 2025 and rolled out progressively, first in the United States. On 8 October 2025 it officially arrived in Italy, together with 35 other languages and almost 50 new countries, bringing global coverage to over 200 territories.

AI Mode vs AI Overview: the differences that matter

One of the most common mistakes is treating AI Mode and AI Overview as synonyms. They are different tools, with different logic and different effects on your site’s visibility. Understanding the difference is the first step towards optimising correctly for both.

Query Fan-Out: the mechanism rewriting SEO

At the heart of Google AI Mode is a technical mechanism called Query Fan-Out. It is the invisible engine that decides how many results are retrieved and how the answer is built. Understanding it is essential because it completely changes the approach to content creation.

Query Fan-Out works like this: when a user asks a complex question, the AI does not treat it as a single query. It breaks it down into several related sub-queries, each exploring a different aspect of the topic. These sub-queries are analysed in parallel, gathering information from several different sources, and then synthesised into a single coherent answer.

The data confirms how central this mechanism is: with the Gemini 3 update of February 2026, pages that correctly capture the fan-out are 161% more likely to be cited than those optimised for the main keyword alone.

It is also worth understanding what sets Query Fan-Out apart from the familiar “People Also Ask” section. The difference is structural: People Also Ask goes deeper vertically, expanding on the same topic. Query Fan-Out expands horizontally, exploring different but connected subject areas and building semantic networks that a human user would take hours to construct manually.

The real impact on organic traffic: the data

The numbers leave no room for optimistic readings for those who fail to adapt. AI Mode’s impact on organic traffic is real, measurable and accelerating.

The most revealing data point is the growing gap between impressions and clicks. A site can see its impressions rise by 49% in Google Search Console and its clicks fall by 30% over the same period. This is not a bug — it is a sign that the site is being shown as a source in AI Mode answers, without the user feeling the need to click through to learn more.

It is important, however, to read the data precisely. The impact is not uniform across all query types:

• Simple informational queries (definitions, “what is”, “how does it work”): maximum impact, with CTR down by up to 61%. The AI answers in full, with no need to click.
• Comparative and decision-making queries (“what is the best X for Y”, “X vs Y”): medium impact. Users read the summary but often want to check the source before making the final decision.
• Transactional and local queries (“buy X”, “marketing agency Rome”): low impact. They remain in the traditional SERP with classic competition.
• Specialist technical queries (content requiring advanced expertise): low impact. Users sense the need to dig deeper at the original source.

The right strategy is not to resist change, but to reorient towards the types of content AI cannot fully replace — and towards the goal of being cited as a source, not just ranking.

6 strategies for optimising for AI Mode

Adapting to Google AI Mode does not mean rewriting everything from scratch. It means adding a layer of logic to existing content — and applying that logic systematically to new content. The strategies below are ordered by short-term impact.

01 — Covering the complete semantic cluster for Query Fan-Out

Because the AI breaks every query down into parallel sub-questions, content that answers a single main keyword is far less likely to be cited than content covering the entire topic cluster. The pillar + cluster structure is no longer just an SEO best practice: it is the minimum requirement for relevance in AI Mode.

• Map the fan-out sub-queries for every important topic: ask ChatGPT or Gemini “if someone asked X, what related questions might they have?” — it is a rough but effective simulation of the real fan-out.
• Structure the main guide with H2/H3 sections dedicated to each sub-query, not just to the main keyword.
• Create linked satellite articles for sub-queries with enough volume to justify a standalone page.
• Use explicit internal links between pillar and cluster pages — the AI also reads links as a signal of topical architecture.

02 — Answer First: a direct answer in the first 40-60 characters of each section

AI Mode extracts chunks of text from each page. The likelihood of extraction depends on how clear the answer is and where it sits: content that answers immediately, in the first sentence of each section, gets extracted. Content that builds a preamble before answering gets skipped.

• Every H2/H3 section must open with the direct answer to the question implied in the heading — not with a preamble, not with a “before answering, it helps to understand that…”
• Use the “Question → Direct answer → Deeper detail” format as a mental template for every informational paragraph.
• Add a TL;DR at the start of every long guide — it is the chunk most likely to be extracted and cited.
• According to Google itself, AI Mode users ask questions 2-3 times longer than average — content must be ready to answer complex questions, not just 2-3 word keywords.

03 — Comparative and decision-making content

Analysis of more than 680 million AI citations shows that comparative content accounts for 32.5% of all citations in generative engines (Profound, 2025). It is the format AI selects most often — and also the most resistant to zero-click syndrome, because users often want to check the source before deciding.

• Crea pagine di confronto esplicite: “X vs Y”, “I migliori X per [scenario specifico]”, “Come scegliere tra A e B”.
• Structure comparisons with clear tables with explicit criteria — AI extracts them easily and cites them with attribution to the source.
• Include explicit recommendations with selection criteria: “Choose X if…, choose Y if…”. AI summaries tend to reproduce exactly this structure.
• Add concrete numerical data to comparisons: they are by far the most citable chunks.

04 — Schema Markup specific to AI Mode

Structured data is the language in which AI systems read your content without interpreting it. Correct markup increases the likelihood that AI will recognise and correctly cite your page as a source.

05 — Enhanced E-E-A-T: verifiable authority

Initial data on AI Mode shows a clear trend: AI tends to cite mainly well-known, trusted brands, favouring sources that demonstrate verifiable expertise. Good content is not enough — AI must be able to verify who wrote it and why that source is authoritative.

• Every article must have an author with a bio, credentials and a link to a verifiable LinkedIn profile. Not “Editorial Staff”, but a real person with a real track record.
• Cite primary sources with links: original studies, government data, academic research. The AI checks citations and favours those who cite authoritative sources.
• Publish proprietary data: internal surveys, (anonymised) client analyses, industry benchmarks. This is the least replicable and most cited content.
• Actively monitor and manage brand mentions on industry sites — external citations are a signal of authority that the AI reads across sources.

06 — Direct channels: reducing dependence on Google

This is the hardest strategy to accept but the most honest to include. If AI Mode structurally reduces referral traffic from Google to websites, the rational response is to build channels that do not depend on Google to bring in users.

• Newsletter: the channel most resilient to the algorithm. A subscriber is yours, not Google’s. 2025 saw 34% growth in B2B newsletter subscriptions in Italy.
• Brand search: the better known your brand is and the more it is searched for directly, the less you depend on informational queries intercepted by AI. The goal is for users to search for “Bliss”, not just “marketing agency Rome”.
• Communities and owned spaces: forums, private Slack groups, events — spaces where interaction happens outside the SERP.
• Video and podcast content: platforms such as YouTube are sources frequently cited by the AI itself — owning YouTube means being present in two channels with a single piece of content.

How to measure visibility in AI Mode

Measurement in AI Mode requires an updated analytics framework. Traditional metrics (average position, aggregate organic CTR) no longer capture reality. The new set of KPIs must combine AI visibility signals with conventional traffic signals.

The integrated strategy: SEO + GEO + AI Mode

Three tools, one logic. SEO, GEO and AI Mode optimisation are not separate disciplines that cannibalise one another: they are layers of the same system. SEO builds the organic ranking that is still the primary selection factor for AI sources. GEO turns that content into something generative engines can cite. AI Mode optimisation adds the specific structure needed for Query Fan-Out and conversational search.

The figure that settles the priorities once and for all: between 76% and 92% of sources cited in AI Overview and AI Mode come from the top 10 organic results (Ahrefs, 2025). Traditional ranking is not dead — it remains the necessary condition. But it is no longer sufficient. Content structure, fan-out coverage and E-E-A-T signals have become equally necessary conditions for turning a good ranking into an AI citation.

Competitive advantage is being built now. The Italian market is still largely unprepared — few sites have updated their content architecture for Query Fan-Out, very few have implemented Speakable markup, and almost none have a system for monitoring AI citations. Those who move methodically in the coming weeks will build a positioning that becomes progressively more expensive to replicate.

FAQs on Google AI Mode

What is Google AI Mode?

Google AI Mode is the Gemini-based conversational search mode built directly into Google Search, live in Italy since 8 October 2025. It turns classic search into a dialogue: users ask questions in natural language, receive concise answers with cited sources and can continue with follow-up questions while keeping the context.

What is the difference between AI Mode and AI Overview?

AI Overview appears in the standard SERP as a summary box for simple queries. AI Mode is a dedicated, conversational mode for complex queries; it uses the Query Fan-Out mechanism and supports multi-step interactions with voice and images. Its impact on organic traffic is deeper and structural.

What is Query Fan-Out?

Query Fan-Out is the mechanism by which AI Mode breaks a complex query down into multiple related sub-queries, analysed in parallel across different sources. Pages that cover the complete semantic cluster — answering several sub-queries within the same content architecture — are 161% more likely to be cited (post-Gemini 3 update data, February 2026).

Does AI Mode reduce my site’s organic traffic?

Yes, on informational queries. Data show a CTR drop of between 34.5% and 61% when AI summaries appear. Only 4.5% of AI Mode sessions lead to a click on the sources. Traffic falls but becomes more qualified: those who arrive are at an advanced decision-making stage. The strategic response is to optimise to be cited as an authoritative source, not just to rank.

How do you measure visibility in Google AI Mode?

Since May 2025 Google Search Console has recorded impressions and clicks from AI Mode. The telling pattern is rising impressions with falling CTR on the same queries: it means presence in AI answers without clicks. Complement GSC monitoring with monthly manual tests on target queries in AI Mode and with tracking of referral traffic from AI sources in GA4.