GEO Insights

Guides, analyses and studies on visibility in AI.

GuidesNovember 10, 2026

Claim your business in ChatGPT (Business Hub)

Since 10 November 2026, OpenAI has let businesses claim their listing in ChatGPT, in beta, under the name Businesses in ChatGPT. From the Business Hub, a verified business updates its information, adds a catalogue with prices and availability, and sees how many times ChatGPT mentioned it, how many times its listing was viewed and how many clicks went to its website. OpenAI opens the process to any eligible business located in a country where ChatGPT is available, Switzerland included; on launch day, access from Switzerland went through a waitlist. The Business Agent, which answers on the business's behalf in ChatGPT, is limited for now to e-commerce and online services in the United States. Everything is free during the beta. According to OpenAI, claiming a listing should not change how often a business appears: recommendations are still drawn from what the web says about it.

15 min
Read article
GuidesOctober 8, 2026

How do ChatGPT and AI assistants cite a company?

ChatGPT and other AI assistants talk about a company in two ways. Either with what the LLM learned during training, its training data. Or with what it searches for on the web at the moment it answers, which is called grounding. On the Swiss questions Repliq tracks, fewer than one answer in ten cites no source at all: web search decides almost everything. Before answering, ChatGPT turns the question into several searches, the fan-out queries, and goes looking for the federal law, the canton, the FRC consumer association or the brand's official site. When the question says Switzerland, the sources turn Swiss. When it does not, the assistant lands on French websites first. A page that clearly states Switzerland, Swiss francs and the year therefore starts with an advantage.

16 min
Read article
InsightsSeptember 3, 2026

AI search in French-speaking Switzerland: what changed

Since April, Repliq Research has been putting the same questions to the four AI assistants we track in French-speaking Switzerland. This summer the gaps between them widened sharply: Google increasingly answers local searches with business listings and now names about a quarter fewer businesses than in spring, ChatGPT shows three times fewer sources than it did in mid-August even though its answers grew 22% longer, and Reddit has all but disappeared from its citations. The other two assistants stayed far steadier. But the most interesting part may be what did not change: roughly one citation in two points to a site the company concerned does not own, and on a question as competitive as which is the best Swiss bank, the most cited page in our corpus belongs to a personal blog.

19 min
Read article
GuidesAugust 22, 2026

When AI gets your company wrong, the error rarely comes from your site

The investment of recent years went into the website: the right information, navigation that leads the user to the right page. AI now reduces all of that to an answer a few lines long, drawing 90 % of its cited pages from third parties, from comparison sites to directories, press and forums. One answer in 14 carries at least one false statement, and every organisation we track has one in circulation. Repliq's Reliability feature compares each measured answer against your official information, then traces the page feeding the error.

7 min
Read article
GuidesAugust 13, 2026

Appearing in an AI Overview: 10 sectors measured

Before trying to appear in an AI Overview, find out whether Google produces one in your sector at all. The rate passes nine searches in ten in online banking and falls below one in two in luxury watchmaking. The format retained changes too: the editorial article leads almost everywhere, by a wide margin, and the directory listing leads in restaurants. The number of sites to work doubles from one sector to another, and so does the observation period needed before judging a result.

6 min
Read article
GuidesAugust 12, 2026

Choosing a GEO agency in Switzerland: six questions

Two GEO offers at the same price can measure very different things: not the same engines, not the same questions, not the same markets. Six questions surface the gap before you sign. They cover the accuracy of the information about your company, what will actually be measured, the role of classic search optimisation, which Swiss sources are targeted, and how long the observation runs. The reference points come from Swiss Atlas, the observatory Repliq runs daily across ten sectors.

10 min
Read article
GuidesAugust 6, 2026

Google AI Overview: when it actually appears

AI Overview is the text Google displays above the links, written by the AI at the moment of the search. Across more than 2,800 questions asked in French-speaking Switzerland, it appears in three cases out of four. It almost always answers “how much does it cost”, almost never “where to buy”: when it is time to choose where to buy, your customers get the links back. The sources it cites are not Google's top results, and they change completely from one national language to another.

9 min
Read article
Case StudiesJune 30, 2026

GEO 2026: how ChatGPT and Perplexity cite you

Your site can rank first on Google and still be absent from ChatGPT or Perplexity answers. That is normal: answer engines do not rank links, they cite a few sources to compose an answer, and each one has its own logic. Repliq recorded the pages that ChatGPT, Google AI and Perplexity cite most in Switzerland, then dissected their content. 4 characteristics come back page after page: a title that echoes the question, a high density of numeric data, a visible update date, an identified author. On top of that comes a major difference between engines: ChatGPT avoids commercial comparison pages but cites reviews (Trustpilot) and official sites, whereas Google and Perplexity readily pick up blogs and comparison sites. This guide names the pages, gives the numbers, and ends with an applicable checklist.

11 min
Read article
InsightsJune 9, 2026

Share of Model: definition, calculation and reliability

Share of Model measures how often a brand appears in answers generated by answer engines (ChatGPT, Google AI Mode, Google AI Overview, Perplexity, Gemini) across a defined set of questions. Its definition is broadly agreed on, its calculation method far less: the choice of denominator, the exclusions and the deduplication rule shift the result by tens of points. Across 8,261 measured answers, the metric proves stable week to week and markets remain fragmented.

15 min
Read article
Case StudiesMay 19, 2026

Swiss hospitality faces a new invisible ranking: the one set by AI

Which Swiss hotels appear in the recommendations of ChatGPT, Gemini and Perplexity, and on what data foundations do these choices rest? This preview summarises the first mapping of AI visibility for the destinations of Geneva, Lucerne, Zermatt and Interlaken. The study analyses each property's share of voice and the weight of authority channels (official websites, Michelin Guide, Wikipedia, consortiums) in algorithmic decisions. The full report, including rankings by stay category and optimisation levers, is available now. Access at the bottom of this page.

13 min
Read article
InsightsMay 18, 2026

Your Brand Now Has an AI Reputation. Are You Measuring It?

AIs do not passively reflect a brand's reputation. They reconstruct it. From customer reviews, articles, forums, comparators, competitor content and sometimes old sources, they produce a synthesis that already influences your customers, candidates, partners and investors. This article walks through five questions to steer this new reputation: what AIs say about you, the sources they use, the queries they trigger, the competitors they associate with your brand, and the alignment between your real strategy and your generated image.

10 min
Read article
Case StudiesMay 7, 2026

From search to recommendation: who answers for your brand in AI?

AI is turning information search into a direct recommendation where users expect arbitration. Our study of 8,800 citations shows that leading Swiss brands today control only 5 to 12% of their own narrative inside these interfaces. This guide unpacks how challengers like rappn and Magic Heidi break through with structured content, and how established brands can reclaim the conversation by becoming reference sources for the algorithms.

9 min
Read article
GuidesDecember 4, 2025

What is GEO? Definition and KPIs in 2026

GEO (Generative Engine Optimization) covers what a company does to be named, cited and accurately described in the answers of AI assistants: ChatGPT, Google AI Mode, Google's AI Overview, Perplexity, Gemini, Claude or Grok. It is measured with four indicators: visibility rate, share of voice, sentiment and position. It complements Google search optimisation without replacing it, because assistants rely mostly on third-party sites and their answers change from one language to another and from one week to the next.

7 min
Read article

Follow what AI answers in Switzerland

New studies, sector analyses and measured cases, with every publication. Repliq tracks ChatGPT, Gemini, Google AI Mode, Perplexity and Grok.

No sales follow-up. Unsubscribe in one click.