Google AI Mode: how to optimise for generative search in 2026

Google AI Mode is the biggest transformation of search since 2010. Not a small update, not an additional feature: a paradigm shift. The classic 10 blue link SERP is giving way to a synthesised answer generated by Gemini that integrates 3-5 cited sources. For anyone doing serious SEO, ignoring this change equals doing email marketing in 2010 while ignoring Gmail's spam filter.

In this guide we explain how AI Mode really works in 2026, how to optimise your content to land in generative answers and why EEAT (Experience, Expertise, Authoritativeness, Trustworthiness) became a technical requirement, not just a recommendation. It's the same framework we apply in SEO projects run by +Click.

#AI Mode: what it really is (and what changes)

AI Mode is Google's new search interface powered by Gemini. When a user queries, instead of seeing the classic link list, they see an AI-written answer synthesising information from multiple sources, citing them with small clickable badges. Below, the traditional organic results remain, but more and more people stop at the generative answer.

Technically, AI Mode uses a fine-tuned version of Gemini with real-time access to Google's index. When you receive a citation, your site is shown as one of the sources that fed the answer. It's no longer a numeric position (1st, 2nd, 3rd), it's a binary inclusion: either you're in the sources or you're not.

The progressive rollout started in the US with SGE (Search Generative Experience) and expanded to Europe with gradual activation and GDPR adaptations. In Italy in 2026 coverage is significant for informational queries, still limited for commercial and transactional ones (Google keeps classic SERP where it knows there's money in play).

#AI Mode vs classic search: 5 practical differences

Understanding where and when AI Mode appears is the first step to adapting strategy. Five concrete differences we see in the Italian market in 2026.

1. Format trigger: AI Mode activates on complex informational queries ("how to optimise the pantry", "what are the symptoms of vitamin D deficiency") and on long-tail multi-intent queries. On short commercial and brand queries Google keeps the classic SERP.
2. Number of cited sources: typically 3-5 links in citations, with different weight. The first cited source receives about 60% of residual clicks, the others split the rest.
3. Type of content rewarded: specific and complete content (1,500-3,000 words), with sub-sections answering sub-questions, formats like "numbered steps", "definition lists", "comparison tables".
4. User behaviour: 47% of Italian users exposed to AI Mode stop at the synthesised answer without clicking. The 53% who click do so only on 1-2 specific sources.
5. Brand visibility: even without clicks, appearing as a cited source maintains brand awareness. Citations work as "social proof snippets" for your brand.

#How to land inside generative answers

Citations in AI Mode aren't random. Gemini selects sources based on a mix of factors that recall classic SEO but with different weights. Let's see what really works, based on what we observe in real projects.

Specificity and topical completeness

Pages that get cited share one trait: they cover the topic completely, with sub-sections answering sub-intents. An article on "how to choose a pellet stove" with sections dedicated to power, installation costs, maintenance, Italian brands and tax deductions beats a generic 800-word article on the "best type of stove". Completeness wins over brevity.

Identifiable author with verifiable authority

Gemini favours content signed by real authors with recognisable web presence. Person schema markup with sameAs to LinkedIn, articles published elsewhere under the same name, dedicated author profile on the site with detailed bio. Without a "traceable" author the content gets downgraded from AI sources. We covered it in the E-E-A-T 2026 guide.

Freshness and documented updates

AI citations favour content with recent datePublished and dateModified. A recently updated page with visible changelogs ("updated May 2026") beats a 2022 page even if well-written. For topics that change over time (technology, ads, laws), periodic updating is non-negotiable.

Primary data, not just opinions

Gemini cites more willingly pages that report primary data: case studies with real numbers, original surveys, own research. When you write an article, try to include at least one data point only you can cite (anonymised internal client data, sector benchmarks). Unique value makes you citation-worthy.

Scannable structure with clear answers

Cited content often has a structure that facilitates extraction: clear H2s and H3s, short paragraphs, tables, numbered lists, inline FAQs. Gemini parses fast and favours texts where the answer is "identifiable" in few tokens. A wall-of-text page is skipped even if it contains the right answer.

#Why EEAT became decisive (and no longer optional)

In classic SEO EEAT was a soft signal: it mattered, but you could rank without explicitly demonstrating it. With AI Mode it became the source-selection filter. Gemini has to choose who to cite and doesn't want to risk citing unreliable content: it looks for authority signals before selecting.

• Experience: content must show direct experience. Phrases like "we tested", "in our projects", "in 11 years of work" are positive signals. Abstract content without proof of direct experience gets discarded.
• Expertise: the author must be competent in the topic. A medical article signed by a generalist blogger is worth less than one signed by an identifiable doctor.
• Authoritativeness: the site must be recognised as authoritative in the sector. Incoming links from sector sites, external citations, presence on vertical directories.
• Trustworthiness: the site must be transparent. HTTPS, privacy policy, visible business address, real contact, affiliations and sponsorships declared.

#Schema markup in the AI era: what Google actually reads

Schema markup became the main way Gemini "understands" your site without having to interpret text. Implementing it well is the difference between being considered and being ignored.

The 6 essential schemas in 2026

1. Organization: with name, logo, sameAs to official social profiles, complete contact information.
2. Person: for each site author, with jobTitle, sameAs LinkedIn/public profiles, alumniOf if relevant, knowsAbout to declare competencies.
3. Article or BlogPosting: for each content, with author reference to Person, datePublished, dateModified, wordCount, mainEntityOfPage.
4. FAQPage: for inline FAQ sections, fundamental to get citations on question-form queries.
5. BreadcrumbList: for navigation, helps Gemini understand site structure.
6. WebSite with SearchAction: to activate sitelinks searchbox in branded results.

Advanced schemas for specific topics

Vertical topics need specific schemas that significantly increase citations. Examples: HowTo for step-by-step guides, Recipe for food content, Product and Offer for ecommerce, LocalBusiness for local activities, Event for events, Course for training. More specific the schema, better Gemini understands the context.

#New metrics: beyond clicks and impressions

Measuring SEO in 2026 with only 2020 metrics is like measuring ad campaigns looking at impressions alone. Metrics to add to the dashboard, alongside the classics.

• AI citation count: how many times your domain appears as a source in AI answers. Tools like Ahrefs, Semrush and Otterly are introducing dedicated metrics.
• Brand mentions in AI: how many times your brand is named in AI answers (with or without link). Pure brand awareness.
• Click-through from citations: what percentage of users exposed to a citation clicks your link. Typically 10-25% on the first two citations, 2-8% on others.
• AI visitor dwell time: those arriving from citations tend to be more qualified (they already know what they want from the answer) and convert better.
• Search Console Performance for AI features: Google started separating AI Overview-related metrics in Search Console in 2025. Check this section weekly.

To integrate these metrics into standard marketing reporting, see the guide on Google Analytics 4 and marketing dashboards, where we explain how to build a cross-channel dashboard that also includes AI traffic.

#What to do today on your site: operational checklist

To not leave the guide theoretical, here's the operational checklist we apply in our SEO projects in 2026. 10 concrete actions to do in the next 30 days.

1. Author profile audit: every post must have an author with dedicated bio, photo, qualifications, link to LinkedIn and public profiles.
2. Person schema implementation for each author, with sameAs to at least 3 verifiable public profiles.
3. Audit of top content: your 20 articles generating most organic traffic in 2025 need updating with refreshed datePublished, new data, recent examples.
4. Adding inline FAQs on every informational article, with correct FAQPage schema.
5. Verify every article has at least one primary data point (internal case, own research, original statistic).
6. "About us" page audit to declare expertise, years of experience, sectors covered.
7. Insert clear citations in your articles: when you cite external sources, do it explicitly with link to the original document.
8. Check every page has title and meta description coherent with intent (no keyword stuffing).
9. Complete Article/BlogPosting schema implementation with all recommended fields.
10. Setup AI citation monitoring via Ahrefs Brand Radar, Semrush AI Search or Otterly.

#Real case: how we adapted a site for AI Mode

To make this concrete, a recent adaptation example. The Appartamenti Mare Sardegna case: bilingual IT/EN platform built in custom Next.js with impeccable technical SEO from day-1, complete LodgingBusiness schema, correct hreflang.

• Complete schema markup: Organization, LodgingBusiness for each apartment, Article for each informational page, FAQPage for area and service questions.
• Identified author: every content signed with Person schema, bio on the author page, links to public profiles.
• Primary data: "What to do in the area" section with specific advice based on direct experience, not copy-paste from TripAdvisor.
• Documented updates: every page has visible dateModified with minimum quarterly update.
• Result: the site is regularly cited as a source in AI Mode answers on queries like "where to sleep in Costa Smeralda", "Sardinia north vacation tips", "sea-view apartments Olbia".

Similar cases with solid technical SEO have generated measurable results: the Parafarmacia Sana case shows how locally specific and author-signed content can emerge in AI searches of healthcare interest for the area.

AI Mode hasn't eliminated SEO, it has drastically raised the bar. The sites surviving in the next 3 years are those investing today in real authorship, primary data and complete schema markup. Those continuing to do keyword stuffing on generic content will disappear.

— Niccolò Giuseppetti, founder +Click

#The 5 most common errors in the AI Mode era

1. Publishing content without identifiable author. Reason number one for being excluded from citations.
2. Stuffing pages with keywords without semantic structure. Gemini doesn't read density, it reads meaning.
3. Not updating old content. A 2021 page without recent dateModified gets demoted.
4. Ignoring schema markup. Without structured data Gemini has to interpret text, and often doesn't pick you.
5. Thinking AI Mode is just "more". It's the new standard, not a separate option.

Checklist to be "AI-ready"

1. Every article has a real author with Person schema and dedicated profile.
2. The 20 most-read 2025 contents have been updated in 2026.
3. All content has complete Article/BlogPosting schema and FAQPage where relevant.
4. Pages include primary data (internal cases, original statistics).
5. Structure has clear H2/H3, short paragraphs, inline FAQs.
6. Organization and Person schemas validated with Rich Results Test.
7. AI citation monitoring configured (Ahrefs Brand Radar or equivalent).
8. Search Console verified and AI Overview segment checked weekly.

#FAQ AI Mode and generative search