Tech for Retail 2025 Workshop: From SEO to GEO – Gaining Visibility in the Era of Generative Engines

Back to blog

Google AI Mode: how it works, which sources it cites and how brands stay visible

GEO

Discover Incremys

The 360° Next Gen SEO Platform

Request a demo
Last updated on

18/11/2026

Chapter 01

Example H2
Example H3
Example H4
Example H5
Example H6

Google AI Mode replaces the results page with a conversation: the user asks a question, reads a written, sourced answer, then follows up without leaving Google. This page is for B2B marketing and SEO leads who want to know what this surface changes for their brand, how to get cited in it and how to check the result.

 

It extends our overview of the AI search engine, which compares how each assistant picks its sources. Here there is a single subject: AI Mode, from the way it builds an answer to the way you measure your presence.

 

What Google AI Mode is

 

Google AI Mode is a conversational search experience built into the search engine. Instead of ranking ten links, Google writes a complete answer from pages it has retrieved, displays links to those pages and invites the user to continue the exchange. It sits alongside the classic results page rather than replacing it.

 

A conversation in place of a results page

 

In a classic search, the user types a keyword, scans the results and opens several pages to piece an answer together. In AI Mode, they describe their need in natural language, with its constraints, and receive an organised summary: an explanation, criteria, sometimes a comparison table, then links. They can then refine the request ("and for a small business?", "compare these two options"), and each follow-up takes the previous exchange into account.

 

Three features set it apart from a general-purpose assistant. It relies on Google's index and on Google's own databases, such as entity records, products and places. It shows links to web pages in support of its answer. And it accepts voice and images as input, which widens the range of search situations well beyond typed text.

 

Availability and access

 

At the time of writing, AI Mode is available in France since summer 2026, like AI Overviews, and availability varies by country and language. Users reach it through a dedicated tab on the results page, or from an AI Overview when they choose to dig deeper. Use of AI engines is no longer marginal: 39% of French people use AI engines for their searches (IPSOS, 2026).

 

For a brand, the consequence is simple. Part of the searches that used to pass through your results pages now takes place in an interface where you only appear if the answer cites you. AI Mode adds to classic search without removing it: the two coexist, and your pages compete on both with the same content. That is why the work described below is an extension of your existing search programme, not a separate project.

 

AI Mode and AI Overviews: two surfaces, two logics

 

AI Mode and AI Overviews rely on the same family of models and the same index, but they do not play the same role in the journey. An AI Overview is a summary placed above the results: the user did not ask for it, and they either read it or scroll past. AI Mode is a choice: the user enters it to work through a question that takes several steps.

 

The table below compares the two surfaces on what matters for your decisions:

 

CriterionAI OverviewAI ModeWhat it means for you
TriggerAutomatic, on some queriesDeliberate, via the tab or a follow-upAI Mode captures more engaged intents
FormShort summary above the linksLong conversation with successive follow-upsMore sub-questions to cover
Typical queryKeyword or simple questionLong question with context and constraintsWork on the question, not just the keyword
SourcesA few links, close to the top resultsMore links, renewed at each follow-upA page can enter or drop out between turns
Classic resultsStill visible underneathAbsent from the screenWithout a citation, no visibility

 

The last row matters most. Under an AI Overview, your links remain on screen: a well-ranked page still has a chance of being clicked. In AI Mode, the list of results disappears from view, and the citation becomes the only form of presence. The effect of overviews on click-through rate, and the criteria for being picked up, are covered in our article on Google AI Overviews and SEO.

 

How AI Mode builds its answer

 

AI Mode does not answer "from memory": it searches, selects, then writes. Understanding that sequence shows at which point a page can be retained, and why it gets set aside.

 

One question split into several searches

 

Faced with a complex question, Google does not run a single search. It breaks the request into sub-questions and runs several searches in parallel, a technique its documentation describes as query fan-out. A question such as "which tool should I choose to track my brand's visibility in AI answers across three markets" can produce searches on types of tool, selection criteria, multi-country tracking and the limits of measurement.

 

This has two effects for you. A page can be retrieved for a sub-question you never targeted as a keyword, provided a section answers it clearly. And a page that only addresses the main query, without the criteria, limits and special cases, leaves the sub-searches to other sources. Coverage of a topic therefore matters more than the repetition of a phrase.

 

How passages and links are selected

 

Retrieved pages are not reused in full. The system retains passages: a definition, a step, a table row, a figure with its source. It then writes the answer and attaches links to the pages those elements came from. Four criteria recur when two passages compete:

 

  • Clarity: the passage keeps its meaning once lifted out of its context.
  • Freshness: dated, maintained information beats an older version.
  • Consistency: what the page says matches what other reliable sources say.
  • Attribution: the author, the brand and the date can be identified.

 

Freshness is not a detail: 70% of pages cited by AI were updated in the last 12 months (The 2026 State of AI Search, 2026). A reference page that nobody revises therefore loses its place as a source, even if its subject has not changed.

 

What AI Mode changes for brand visibility

 

AI Mode shifts two things: the shape of demand and the value of a visit. Both can be measured, and neither is handled by a simple tag adjustment.

 

Longer questions, more precise intents

 

In a conversation, users no longer squeeze their need into three words. They state their sector, their size, their budget and their constraints, then refine through follow-ups. Your unit of work changes: you are no longer aiming at an isolated keyword, but at a family of questions asked by the same type of reader.

 

That changes how you prepare a page. Start from the persona and list the questions they ask at each stage: framing the need, comparing options, removing an objection, checking a condition. A page that answers this sequence, section by section, has more opportunities to be picked up than a page built around a single phrase. Follow-ups often carry the objections and the edge cases: those are what you need to write, not just the definition.

 

Fewer visits, but more advanced visitors

 

A complete answer satisfies part of the searches without a visit. The trend has grown with generated answers: zero-click searches have multiplied by 2.5 since AI Overviews (State of AI Search, 2025). Expect simple informational queries to bring fewer sessions.

 

There is a counterpart. A reader who clicks from an AI answer has already read a summary and comes looking for what it does not provide. Visitors from AI answers are 4.4 times more qualified than those from classic search (Squid Impact, 2025). The decision that follows concerns your pages: those that must be cited (definitions, criteria, methods) and those that must convert the visit (detailed comparison, demonstration, proof). Judging the first group on sessions alone underestimates what it contributes.

 

How to get cited in Google AI Mode

 

No markup "declares" a page to AI Mode. Google says so in its documentation: search best practice remains the foundation, and a page must be indexed and eligible to be shown with a snippet in order to be reused. The work falls into three workstreams, to be handled in this order.

 

Check access and eligibility

 

Without crawling and indexing, no citation is possible. AI Mode uses the Google Search index, so it is Googlebot that needs to reach your pages, not a separate crawler. Check four points before touching the content:

 

  • Crawling: no robots.txt rule and no block at hosting level prevents Googlebot from reading strategic pages.
  • Indexing: no accidental noindex tag, consistent canonical tags, no duplicates.
  • Snippets: the nosnippet, data-nosnippet and max-snippet directives limit what Google may reuse, AI features included; make sure they are not applied by default.
  • Rendering: the main content is present in the HTML Google sees, which the URL inspection tool confirms.

 

These checks are carried out page by page in Google Search Console, on the templates that carry your acquisition.

 

Answer the question and its follow-ups

 

The second workstream is editorial. Each section must be reusable on its own: a complete answer sentence at the top, then verifiable elements, then conditions and limits. Structure the page so that it follows the conversation:

 

  • The main question: a direct answer in two or three sentences, placed first.
  • The criteria: a list or table that allows comparison without rereading the whole text.
  • The special cases: by size, sector, budget or country.
  • The objections: what does not work, the limits, the common mistakes.
  • The next step: what the reader should do after reading.

 

The formats that lend themselves to extraction, and the protocol for checking them, are described in our article on AI-optimised content. Avoid multiplying near-identical pages for every variant of a question: one reference page per intent, complemented by satellite pages with distinct angles, is reused more readily than ten pages competing with one another.

 

Make the brand verifiable beyond its own site

 

The third workstream happens elsewhere than on your pages. A generated answer cross-checks its sources: what the web says about your brand must match what you publish. Company name, description of the offer, leadership and key figures should be identical on your site, on your public profiles and on the sites that talk about you.

 

Mentions on sources the engines already reuse, such as trade media, comparison sites and community platforms, weigh in the selection. Structured data helps describe who is publishing and what the page is about, without ensuring that it is picked up. Consistency of entities counts for more than the volume of markup.

 

A simple check gives you a starting point. Ask AI Mode to describe your company and your category, then compare the answer with your own positioning. Where the answer is vague, outdated or attributes your features to someone else, you have found the gap to close first: an "about" page that states plainly who you are, product pages that use the same vocabulary as your public profiles, and third-party pages that describe you in the same terms. Repeat the exercise after each round of corrections, with the same wording.

 

Measuring your presence in AI Mode

 

Google does not provide a report dedicated to AI Mode. Measurement therefore combines two readings: what your tools record, and what you observe yourself.

 

What Search Console records

 

At the time of writing, Search Console counts impressions and clicks from AI Mode in the Performance report, together with those from AI Overviews and classic results, without separating them. You cannot isolate an "AI Mode" line. Three signals can nevertheless be read, page by page:

 

  • Impressions up without clicks rising: the page is seen in answers that satisfy the request.
  • Click-through rate down at a stable position: a generated answer captures attention before your links.
  • Longer queries in the report: conversational wording appears among the page's queries.

 

Compare each page with itself, before and after a change. A gap is a correlation over time, not proof of cause: seasonality and competition also play a part.

 

The question-based reading

 

Presence in answers cannot be read in a Google dashboard: it has to be recorded. Build a fixed set of questions per persona, worded as a reader would ask them, and submit them to AI Mode at regular intervals. For each question, note one of four states:

 

  • Brand named in the answer.
  • Domain cited through a link in the answer.
  • Domain as a source, listed without being mentioned in the text.
  • Absent.

 

An answer varies from one session to the next: repeat the reading before drawing conclusions, and follow the trend rather than a single result. Read AI Mode separately from other engines, without averaging. The full set of indicators, from citation per question to conversions, is defined in our article on GEO KPIs.

 

Which pages to work on first

 

Not every page deserves the same effort. Sort them by the intent they serve and the value they carry:

 

Type of pageExposure to AI ModeObjectivePriority action
Definition, conceptHigh: the answer is often enoughBe citedDirect answer first, stable definition, update date
Guide, methodHighBe cited and keep the visitNumbered steps, validation criteria, edge cases
Comparison, choiceMediumInfluence the shortlistTable of measurable criteria, explicit assumptions
Study, proprietary dataMediumBecome the sourceMethod, scope, source and year for each figure
Offer, demonstrationLowConvert the visitProof, conditions, answers to objections

 

Start with pages that combine many impressions and an intent the answer can summarise: their visibility is decided by citation. Then handle the choice pages, where a citation confirms your place in the buyer's shortlist. Further benchmarks on these behaviours are gathered in our GEO statistics.

 

Whatever the order, set a date for a second reading. A page reworked for extraction usually shows an effect within a few weeks on the questions it answers; if nothing moves over that period, the cause lies elsewhere, most often in access or in the brand's presence beyond its own site.

 

FAQ on Google AI Mode

 

What is Google AI Mode?

 

Google AI Mode is a conversational search experience built into the search engine. It writes an answer from pages retrieved in Google's index, displays links to its sources and accepts follow-up questions. It adds to the classic results page without replacing it.

 

Where is Google AI Mode available?

 

Availability varies by country and language, and it changes over time. At the time of writing, AI Mode is available in France since summer 2026, like AI Overviews. Users reach it through a tab on the results page or from an AI Overview.

 

What is the difference between AI Mode and AI Overviews?

 

An AI Overview is a summary displayed automatically above the results, which stay visible. AI Mode is a conversation the user chooses to open: the list of links disappears, the answer is longer and the sources change with each follow-up. In AI Mode, only a citation makes a page visible.

 

How do you appear in Google AI Mode?

 

A page must be indexed by Google, eligible to be shown with a snippet, and answer the question and its follow-ups clearly. Put a direct answer at the top of each section, add criteria, special cases and limits, and date your updates. No specific markup is enough to obtain a citation.

 

Do you need a special crawler or file for AI Mode?

 

No. AI Mode relies on the Google Search index: Googlebot crawls your pages. No dedicated file is required. The nosnippet and max-snippet directives do, however, limit what Google may reuse, including in its AI features.

 

Can you measure traffic from AI Mode?

 

Not separately. Search Console counts AI Mode impressions and clicks in the Performance report, without distinguishing them from other results. Measurement relies on comparing each page before and after, and on a regular reading of brand presence over a fixed set of questions.

 

Continue reading

 

Depending on what you need to decide now:

 

  • You are starting out: if you have neither a set of questions nor a presence reading yet, begin with the overall method, from audit to measurement → Generative Engine Optimization.
  • You are also targeting assistants outside Google: if your buyers use ChatGPT, selection runs through another index and other crawlers → ChatGPT SEO.
  • Your pages are cited but the brand is not named: if engines reuse your content without attributing the answer to you, the subject is authority beyond your site → backlinks and AI.
  • You need to know where to focus: if you are hesitating between several topics, the Incremys opportunity analysis module separates keywords, which carry a SERP and an AI Overview, from the questions to address per persona, and shows which to prioritise.

Discover other items

See all

Next-Gen GEO/SEO starts here

Complete the form so we can contact you.

The new generation of SEO
is on!

Thank you for your request, we will get back to you as soon as possible.

Oops! Something went wrong while submitting the form.