What Car Brands Does AI Rank as Strongest? ... 'Sorento, Ioniq 5, BMW, Encar.com, Lotte Rental Car'

AIBIX lab said in a company announcement that it conducted an analysis across 5 automobile-related categories, 128 models and brands in total. The 4 generative AI systems analyzed were ChatGPT, Gemini, Perplexity, and Claude, and the analysis covered their car-related responses. The categories analyzed were domestic internal-combustion SUVs, EVs, imported cars, used-car trading platforms, and rental car/car-sharing.

The analysis found that the overall No. 1 brands by category were Kia's Sorento, Hyundai's Ioniq 5, BMW, Encar.com, and Lotte Rental Car. The AIBIX scores for the No. 1 spots were 54.8 for Sorento, 46.5 for Ioniq 5, 65.6 for BMW, 65.2 for Encar.com, and 63.7 for Lotte Rental Car. The 5 No. 1 models and brands also appeared in more than 60% of responses in each category.

The analysis concluded that there were differences between the overall rankings and the order in which items were first mentioned in responses. It also found that the first-mentioned item changed depending on the question conditions. In addition, Lotte Rental Car was the only brand that ranked No. 1 across all 4 AI systems.

On a vehicle-model basis, the analysis of domestic internal-combustion SUV and EV categories showed the strength of Hyundai Motor and Kia. Hyundai Motor and Kia had high market shares in Korea's completed-vehicle market, and the top 6 domestic SUVs were all Hyundai Motor and Kia models. The top 5 EVs were also all Hyundai Motor and Kia models, and 8 of the top 10 EVs were Hyundai Motor and Kia models.

However, the overall rankings differed across AI systems in most categories. For some models and platforms, No. 1 and No. 2 were reversed depending on the AI. Ioniq 5 ranked No. 1 in ChatGPT, Gemini, and Perplexity, while Encar.com ranked No. 1 in Gemini, Perplexity, and Claude.

Sorento ranked No. 1 in ChatGPT and Perplexity, but No. 2 in Gemini and Claude. BMW ranked No. 1 in Claude and No. 2 in ChatGPT, Gemini, and Perplexity.

Meanwhile, in the rental car/car-sharing segment, Lotte Rental Car ranked No. 1 across all 4 AI systems. Lotte Rental Car appeared in 79.2% of responses in the rental car/car-sharing category, and its share of first mentions was 51.7%.

In some categories, the overall ranking did not match the first-mention ranking. The only category in which the overall No. 1 also topped first mentions on its own was rental car/car-sharing. In the used-car trading platform category, the overall No. 1 was Encar.com and K Car, and the top share for first mentions was also shared by Encar.com at 29.2% and K Car at 29.2%.

In domestic internal-combustion SUVs, Sorento was the overall No. 1, but Palisade was No. 1 in first mentions at 22.5%. Sorento's first-mention share was 17.5%, and Palisade ranked No. 3 overall. In EVs, Ioniq 5 was the overall No. 1, but Ioniq 6 was No. 1 in first mentions at 18.3%.

In imported cars, the overall No. 1 also differed from the first-mention No. 1. BMW was the overall No. 1 in imported cars, while Lexus was No. 1 in first mentions at 21.7%. This suggests that the top choice in an overall evaluation and the brand consumers mention first can be different.

When questions specified purchase conditions in greater detail, the leading recommendation changed. In questions about fuel economy for domestic SUVs, the first-mentioned model was Niro, while in questions about parking convenience for domestic SUVs, it was Casper. In questions about ride comfort for domestic SUVs, Genesis GV80 was the first-mentioned model, showing that the model mentioned first varies depending on criteria such as fuel economy, parking convenience, and ride comfort.

The recommendation results also varied depending on the topic and category of the question. For fast-charging priority among EVs, the recommended model was Ioniq 5, while for driving range on a single charge, the recommended model was Ioniq 6. For family-use interior space, the top recommended model was EV9.

In rental car/car-sharing, Carmore was mentioned first in questions about low-cost rental cars at travel destinations. In some questions about monthly long-term rental costs, SK Rent-a-Car was mentioned first.

There were also category-based differences in the sources cited by AI responses. The share of brand official websites among all citations was 27.1% in the used-car trading platform category, the highest among the 5 categories. The share of brand official websites among all citations was 11.5% in rental car/car-sharing. Half of the official-site citations in rental car/car-sharing came from the Lotte Rental Car website and direct booking site.

By contrast, in the vehicle categories, the share of manufacturer official-homepage citations was low or not confirmed. The share of manufacturer official-homepage citations in imported cars was 2.8%, and in EVs it was 1.2%. No direct citations from manufacturer official homepages were confirmed in domestic SUVs.

The analysis found that the citation sources for car-related responses leaned more toward outside content such as news articles, blogs, and portal news than toward manufacturer official sites. In some responses, former company names also appeared.

Cases were also confirmed in which former company names appeared alongside mentions of specific models. In half of the responses mentioning Rexton and Torres, the term 'Ssangyong' also appeared.

In the imported-car category, the share of news article citations was especially high. Citation-source shares in the imported-car category were 36.1% for news articles, 15.8% for blogs, 5.9% for portal news, and 5.1% for YouTube, with news articles plus portal news totaling 42.0%. A similar source pattern appeared in the EV category. Citation-source shares in the EV category were 28.8% for news articles, 14.7% for blogs, and 9.4% for portal news, while public-agency materials such as the Integrated Zero-Emission Vehicle portal accounted for 10.5% of citations in ChatGPT's EV-related responses.

Kim Jun-hyeon, CEO of AIBIX lab, said based on the analysis of the 5 car categories that models and brands frequently recommended may not match the subjects declared at the beginning of responses. He also said that when conditions such as fuel economy, parking, driving range, space, and price are specified more concretely, the model or brand AI presents first can change.

Source: TECHWORLD · Lee Gwang-jae
Original: https://www.epnc.co.kr/news/articleView.html?idxno=407851

References

• https://www.epnc.co.kr/news/articleView.html?idxno=407851