The Auto.ru platform has launched a beta version of the intelligent assistant "Auto.ru AI," which helps users select cars in a dialogue format. The tool is based on a language model trained to work with a large array of structured and textual data from the automotive market.
Unlike classic filters, where the user must specify exact parameters in advance, the new system allows requests to be formulated in free form. The algorithm interprets the text, clarifies the criteria, and matches them with the listings database, model specifications, and user reviews.
Technologically, the solution combines several data sources:
- Offer catalog
- Historical service statistics
- Information on the actual cost of ownership
- Hundreds of thousands of car owner reviews from the "Logbook" section
This data undergoes preliminary processing and structuring, after which it is taken into account when generating recommendations.
A distinctive feature of the approach is the analysis not only of "hard" parameters, such as engine power or body type, but also qualitative characteristics: comfort, reliability, and liquidity on the secondary market. Some conclusions are formed based on the analysis of owner reviews: the system gathers their opinions and consolidates them by the main parameters of vehicle operation.
Two years ago, "Auto.ru AI" began to be used for the automatic generation of listing texts and brief summaries of reports on used cars. Since the summer of 2025, the algorithms have also started to systematize owner reviews across 17 parameters, highlighting the most frequently mentioned features of the models.
The developers note that the tool is operating in test mode. In the near future, they plan to expand interaction scenarios and improve the accuracy of query interpretation.