GEO (Generative Engine Optimization), which AIxH also practises under the name GSO, covers the editorial and technical practices that raise the probability of a piece of content being cited in the answers of generative AI: ChatGPT, Gemini, Perplexity and Google's AI summaries. The subject is no longer theoretical: AI summaries have been live in France since 22 July 2026, and two successive Ahrefs studies measure a drop in organic clicks on the first position of 34.5% in March 2025 and then 58% in December 2025 when an AI summary is shown. This guide answers the four questions business leaders ask us most often: what GEO is, why classic SEO is no longer enough, how to check whether your company appears in ChatGPT and Gemini answers, and how to improve that visibility.
What is GEO (Generative Engine Optimization)?
GEO covers the practices that raise the probability of being cited in an AI-generated answer. The discipline has an academic basis: the founding study, published by researchers from Princeton, Georgia Tech and IIT Delhi in the proceedings of KDD 2024 (Aggarwal et al.), measures the real effect of editorial techniques on the probability of citation. Three levers stand out: citing reliable sources, including statistics and figures, and writing dense, assertive sentences that can be lifted as they are, with measured visibility gains of up to 41% for some methods. Keyword stuffing has no measured positive effect, and sometimes a negative one. These gains remain orders of magnitude measured in a laboratory, on a protocol where a handful of sources compete, and the study predates the general rollout of AI summaries: they cannot be transposed as they stand to a real competitive market. Its hierarchy still holds, though: the synthesis published in 2026 by Cyrus Shepard, drawn from 54 studies, patents and experiments, finds the same editorial levers at the top of the AI citation factors.
Around this base, the consulting literature adds practices that are widely repeated but less rigorously validated: structuring content as questions and answers, answering directly in the first sentence of each section, marking up FAQs in JSON-LD, keeping content visibly fresh, and keeping pages readable without client-side JavaScript. They have one merit: all of them also strengthen the SEO foundation, and none of them replaces it.
One nuance deserves attention before you invest: being cited does not always mean being represented correctly. In March 2025, the Tow Center for Digital Journalism (Columbia Journalism Review) tested eight generative search engines on excerpts from news articles: more than 60% of the answers misidentified the source, its headline, its publisher, its date or its URL, with wide variations from one tool to the next, from 37% of errors for Perplexity to 67% for ChatGPT Search, and almost never any admission of uncertainty. Tracking what AI engines actually say about your brand therefore matters as much as the volume of citations.
Why do organic clicks fall when an AI summary is displayed?
Because the answer appears above the first link. On 300,000 keywords and Search Console data, Ahrefs measured in December 2025 a click-through rate of 1.6% on the first classic result when an AI overview is shown, against 3.9% without one, 58% less; the same team had already measured a 34.5% gap in March 2025. Pew Research Center, on the real browsing of 900 American adults (68,879 searches, March 2025), records a click on a classic result in 8% of visits with an AI summary against 15% without, and only 1% of clicks on a source cited inside the summary.
The strongest evidence of a causal link comes from a randomised field experiment (Agarwal and Sen, April 2026): a browser extension hid or displayed Google's AI summaries for 1,065 American participants. When the summary is shown, outbound organic clicks fall by 39.8% and searches without any click rise by 34.5%, with no effect on sponsored clicks. This work is still a working paper that has not yet been peer reviewed: a strong signal, short of definitive proof.
The decline is not linear, however. Seer Interactive, on 5.47 million tracked queries, observed a floor of 1.3% click-through rate on queries with an AI overview in December 2025, then a rise to 2.4% in February 2026, which the agency describes as a plateau rather than a recovery. Google, through its Vice President of Search Liz Reid, has maintained since August 2025 that the total volume of organic clicks remains relatively stable and that click quality is rising: this position has never been backed by verifiable figures and comes from a party with an obvious conflict of interest; it deserves to be reported, and it should not be taken as a demonstrated counter-argument. Zero-click, finally, predates generative AI: Rand Fishkin has documented it since 2019, and measures in 2026 that 68% of American Google searches end without any click to a website.
Why is classic SEO no longer enough, while remaining the foundation?
The mechanism changes in nature: a classic engine ranks whole pages, a generative engine extracts passages and recomposes an answer. The foundation itself does not change. The official guide published by Google in May 2026 presents optimisation for its generative features as a form of SEO, and warns against techniques sold as indispensable when they are not: no llms.txt file or specific machine-readable markup is required to appear in its generative results, contrary to what many commercial GEO guides claim. We devoted an entire opinion piece to sorting the tactics that are sold from the tactics that are measured.
Ranking well helps, but it no longer guarantees a citation. In March 2026, Ahrefs measured that only 38% of the pages cited in Google's AI overviews also appeared in the top 10 of classic results, against 76% in July 2025: generative engines break the query down into sub-queries, the fan-out, and fetch pages ranked well beyond the first page. This is the double requirement of the moment: without an SEO foundation, no access to extraction; without editorial work designed for citation, no presence in the answer. The differences and bridges between the two disciplines are detailed in our dedicated guide.
Where does the French-speaking market stand, and what do we know about Luxembourg?
Since 22 July 2026, Google's AI summaries and AI Mode have been live in France: every Luxembourg or international company with a French-speaking audience is exposed to the same mechanisms measured on English-speaking markets since 2024.
On the AI search behaviour of Luxembourg consumers and B2B buyers, no identified public study provides a measurement to date, and that gap deserves to be flagged openly; filling it by extrapolation would be a mistake. The available data covers adoption inside companies: according to Eurostat (December 2025), 33.61% of Luxembourg companies use AI technologies, against an average of 20% in the European Union, and the 2026 barometer from FEDIL, the Luxembourg AI Factory and Luxinnovation finds that 59% of organisations have deployed generative AI at scale or in production, a shift we have documented. An economy that adopts AI this quickly inside its own walls has little reason to ignore it in its searches: this is a reasonable inference, and we present it as such; it is not measured data.
How do you know whether your company appears in ChatGPT and Gemini answers?
The most reliable method is still manual: ask ChatGPT, Gemini and Perplexity the questions a potential client would ask about your sector, your city or your products, then note whether your company appears, from what angle, and among which cited sources. Repeating the exercise across several phrasings is essential: generative answers vary from one session to the next, unlike a Google ranking, which is stable for a given query. Presence in Google's AI summaries is checked the same way, query by query, since Google offers no official tool for tracking citations in its summaries.
The structured version of this exercise is a panel of queries replayed at a fixed cadence, month after month, which distinguishes citation from recommendation: being mentioned in an answer differs from being named as the right choice. AI citation tracking tools are emerging, but none of them yet offers the coverage or the reliability of a classic SEO rank tracker. This is precisely what AIxH's free GSO audit covers: we replay this panel for your brand and your target queries, and you leave with a factual picture of your visibility in AI answers.
How do you improve your visibility on ChatGPT, Gemini and Google's AI summaries?
1. Consolidate the SEO foundation first
Technical health, domain authority, content structure: this is a prerequisite. Generative engines rely on the same crawling and indexing foundations as classic search.
2. Structure content as questions and answers
A direct answer in the first sentence of each section, supported by figures and named sources instead of vague wording: this is the lever best validated by academic research.
3. Add markup and keep pages readable without JavaScript
FAQ markup in JSON-LD on strategic pages, and content that can be read without client-side JavaScript, make extraction easier. Markup alone creates no visibility: it supports content that answers.
4. Do not pay for prerequisites that are not prerequisites
The llms.txt file is not necessary to appear in Google's generative results: its official documentation says so explicitly, whatever many sales pitches claim.
5. Track citations month after month
A panel of queries replayed at regular intervals on ChatGPT, Gemini, Perplexity and Google's AI summaries, where a single ranking checked once tells you little: generative visibility is measured over time.
6. Keep the headline figures in their place
The double-digit gains quoted in the marketing press come from laboratory protocols: a useful order of magnitude for setting priorities, with no guarantee on your own market.
AIxH's view
AIxH's SEO & GSO practice treats these two projects as one: consolidating the foundation we build as an SEO agency in Luxembourg, then building presence in generative answers, which we track as a GEO agency in Luxembourg, without setting the two disciplines against each other, since they share the same technical and editorial base. If you want to know where your company appears, or fails to appear, in the answers of ChatGPT, Gemini, Perplexity and Google's AI summaries, AIxH offers a free GSO audit: simply fill in the contact form to request it. You leave with a picture of your AI visibility and the priorities for improving it, before any commitment.