I'm trying to buy an apartment in Seoul. I'd like the price to be 1 billion won or less. At least two bedrooms, and more than 300 households in the complex. I'd like it to be within an hour by public transportation from Gwanghwamun. Find listings that meet these conditions.
After pounding the pavement to check listings in person and doing online sleuthing on maps and social media (SNS), the era of artificial intelligence (AI) field surveys for real estate has arrived. Real estate information platform KB Real Estate and Naver Pay Real Estate have each rolled out services that use AI to find the listings you want. I tried KB Real Estate's "Home-finding AI" and Naver Pay Real Estate's "AI Home Finder."
First, I opened the KB Real Estate application (app) and tapped the Home-finding AI button next to the top search bar. After I entered the conditions I wanted, five listings appeared about 10 seconds later. The information on the apartment complexes and the specific listings, such as price, size, and floor, was organized in a table so I could see it at a glance.
What stood out as a differentiator was the "AI briefing." It took a bit longer to generate, but for listings I was interested in, I could tap "see more" to view details. It combined detailed information, including location notes entered by a licensed real estate agent, with KB's statistical data, and it felt like having an expert next to me summarizing and explaining the listing.
One drawback was that two of the five listings suggested by Home-finding AI were apartments in Namdong District, Incheon, and Siheung, Gyeonggi, not in Seoul. They failed to meet the premise of being "within the Seoul area." The fact that it suggested only five listings at a time was also a bit disappointing. While its natural language understanding is a little lacking, it felt like an assistant that neatly organizes and delivers the needed information.
When I entered the same query through AI Home Finder in the Naver Pay app, it presented eight listings, three more than the other. One listing was in Goyang, Gyeonggi, but it was near Eunpyeong District, Seoul, so it was effectively in the Seoul sphere. The rest were all in Seoul, so in terms of distance it better met the conditions I wanted than KB Real Estate. Naver Pay went further by offering more listing information through a listings map button.
As the first in the industry to apply a large language model (LLM), AI Home Finder looked similar to major AI chatbots like ChatGPT. It could interpret natural language such as "Find a home in Gwanak District, Seoul, for a newlywed couple," rather than strictly specific criteria, and after it found listings, if I added a condition like "Show only apartments," it enabled continuous searching by combining it with the previously desired conditions.
Perhaps because of this, the number of opportunities to ask questions to find a home was twice as many at KB Real Estate as at Naver Pay Real Estate in terms of count. Naver Pay provides five free question credits every Monday. In other words, the number of questions possible in a week is limited to five. Given that the home search process requires comparing and revising multiple conditions and areas, actual users may find this somewhat insufficient. By contrast, KB Real Estate allows 10 new questions every day.
Both services are still offered as beta services. For now, AI field surveys appear most useful as an auxiliary tool to shortlist properties to visit and prepare basic materials to verify on site. Interest, however, is high. The number of listings searched through Naver Pay's AI Home Finder reaches about 10,000 a day, and KB Real Estate's Home-finding AI is said to be used by hundreds of people daily.
A real estate industry official said, "In the past, platforms with more extensive listing information won people's choices, but now, amid this flood of information, the accuracy of AI in pinpointing the 'one right home' needed by the buyer is expected to determine a platform's success or failure."