--- title: "Family cars in Slovakia. Which brands AI recommends to families." author: "Peter Vidovič (GEO strategist)" published: 2026-08-01 url: https://www.petervidovic.com/en/ai-pulse/family-cars/ translation_of: https://www.petervidovic.com/sk/ai-sondy/rodinne-auta/ language: en --- # Family cars in Slovakia. Which brands AI recommends to families. > 39 questions from parents choosing a car, three AI tools, eight runs per question. Which brands and models appear in the answer, how stably, and where AI draws its information from. - Category: Family cars - Measured: August 2026 - AI tools: ChatGPT, Gemini, Google AI Mode - Questions: 39 - Runs per question: 8 - Answers evaluated: 936 - Different brands in answers: 62 ## Key finding: Which car brand does AI recommend to families most often? Škoda. It appeared in 66% of answers; the true value lies between 56 and 74%. It is followed by Volkswagen (63%) and Hyundai (54%). ## Conclusions 1. **The stable core consists of only two brands.** Škoda and Volkswagen appear in all eight runs. 2. **A strong brand fragments into models.** Volkswagen appeared in 63% of answers, but its strongest model, the Touran, in only 21%. 3. **Being among the sources is not enough.** In 30% of cases AI drew on the brand website and still did not mention the brand in its answer. ## Who AI mentions most often Visibility is the share of answers in which a brand appears. Across 936 answers, 62 different brands and 565 different names of brands and models came up. **Brand visibility in AI answers** | Item | Value | Interval | | --- | --- | --- | | Škoda | 66% | 56%–74% | | Volkswagen | 63% | | | Hyundai | 54% | | | Toyota | 52% | | | Kia | 52% | | | Ford | 43% | | | Citroën | 42% | | | Peugeot | 41% | | Source: ChatGPT, Gemini, Google AI Mode · 936 answers · 8/2026 ## Stable core and flickering tail Stability shows in how many of the eight runs of the same question a brand appeared. - **Stable core (8 out of 8 runs):** Škoda, Volkswagen - **On the edge (5 to 7 out of 8):** Hyundai, Citroën - **Flickering tail (1 to 4 out of 8):** Toyota, Kia, Ford, Peugeot ## Models Models show how a brand’s visibility gets divided. Volkswagen appeared in 63% of answers, but its mentions were spread across 42 names of models and variants, so the strongest, the Touran, has only 21%. Dacia appeared in 39% of answers but has only 14 names, so its Jogger reached 26%. **Model visibility in AI answers** | Item | Value | Interval | | --- | --- | --- | | Škoda Kodiaq | 31% | 23%–41% | | Dacia Jogger | 26% | | | Citroën Berlingo | 25% | | | Volkswagen Touran | 21% | | | Hyundai Tucson | 21% | | | Peugeot 5008 | 21% | | | Peugeot Rifter | 20% | | | Volkswagen Sharan | 18% | | Source: 936 answers · 8/2026 ## By tool Each tool received all 39 questions, each one eight times, i.e. 312 answers per tool. Google AI Mode concentrates its answers on a single brand more strongly than the other tools. | Tool | 1st place | 2nd place | 3rd place | | --- | --- | --- | --- | | ChatGPT | Volkswagen 61% | Škoda 56% | Ford 50% | | Gemini | Hyundai 71% | Volkswagen 67% | Škoda 65% | | Google AI Mode | Škoda 77% | Volkswagen 61% | Hyundai 52% | ## Mention and citation A mention is the brand name directly in the text of the answer. A citation is a link in the list of sources. For a brand, the mention matters more; only a fraction of people click on citations. - **59%** of people plan to visit the brand website after a mention [1] - **25.1%** of mentions occur without a citation [2] - **~1%** of visitors click on a citation [3] - **61.7%** of citations are “silent”: the brand does not appear in the answer [2] In this AI Pulse study, in 30% of cases AI drew on the brand website and still did not mention the brand in its answer. ## Where AI drew its information from A ranking of websites, not brands: the share of answers in which AI cited a given website. 584 different sources in total. **Most cited sources** | Item | Value | Type | | --- | --- | --- | | autofilter.sk | 43% | review site | | finalcd.sk | 19% | commercial website | | autoviny.sk | 18% | media | | modrykonik.sk | 17% | community | | highmotor.com | 16% | media | | autorro.sk | 16% | commercial website | | carvago.com | 15% | marketplace | | autobazar.eu | 14% | marketplace | | pravda.sk | 14% | media | | sme.sk | 14% | media | Source: 584 cited sources · 8/2026 ## Where AI found people’s experiences Websites with consumer reviews and discussions, and their rank among all 584 sources. Modrý koník is the only Slovak community platform among the main sources. **Sources of consumer experiences** | Item | Value | | --- | --- | | modrykonik.sk (4th) | 17% | | reddit.com (11th) | 11% | | youtube.com (16th) | 9% | | facebook.com (38th) | 4% | Source: rank among 584 sources · 8/2026 ## Methodology We did not make up the questions. They come from discussions on Modrý koník: we extracted the real questions of parents choosing a car, semantically matched them with the category description and removed brand names from them. - 39 questions × 3 tools × 8 runs = 936 evaluated answers - ChatGPT, Gemini and Google AI Mode, anonymously and without logging in - eight runs per question, because AI answers slightly differently each time [4] - stability bands are evaluated within a single tool: stable core 8 out of 8, on the edge 5 to 7, flickering tail 1 to 4 - the measurement took place in August 2026 and is a snapshot of the market at the time of measurement > **Statistical note** > > Runs of the same question are correlated, so uncertainty is calculated at the question level, not binomially. For Škoda, the DEFF is 9.2, the effective sample is approximately 100 independent measurements and the ICC is 0.35. Ranges are calculated using the Wilson method on the effective sample and are reported only for brands that appeared in at least five different questions. They capture variability between questions, not selection bias or errors in recognising the brand in the text, and they are not a test of the difference between brands. ## Sources 1. [Fractl, survey of behaviour after a brand mention in an AI answer](https://www.frac.tl/ai-search-study/) 2. [Semrush and Kevin Indig, study of mentions and citations in AI answers](https://www.semrush.com/blog/the-ghost-citations-study/) 3. [Pew Research Center, measurement of clicks on sources in AI answers](https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/) 4. [Don't Measure Once (2026), study of repeated measurement of AI answers](https://arxiv.org/abs/2604.07585) PDF: https://www.modrykonik.com/files/ai-sonda-modrykonik-rodinne-auta-august-2026.pdf --- Author: Peter Vidovič, GEO strategist. https://www.petervidovic.com/en/