An increasing number of crypto users inquire with the AI chat tool before opening an account about which exchange to use. A new study shows that such recommendations do not correspond to the actual trading volumes of the exchanges; in multiple mainstream AI engines, Kraken is recommended more frequently than Binance.
Test results of five AI engines
The research covered ChatGPT, Perplexity, Gemini, Claude, and Google AI Mode. The tests did not directly search for brand names but simulated common questions that users might have before opening an account, such as which cryptocurrency exchange is more suitable to use.
A total of 20 buyer-style questions were set up in the study, with each question being tested three times in each engine, resulting in approximately 1000 observations in total. The transaction volume data was taken from CoinGecko on the day of measurement.
The results show that Kraken ranks first overall and leads other exchanges in ChatGPT, Perplexity, and Gemini. Coinbase ranks second and appears the most frequently in Claude, with a total of 83 mentions. Binance comes in third place, although its daily spot trading volume is about 9 billion US dollars, which is approximately 5 times that of Coinbase.
Trading volume is not directly converted into recommendations.
This study reflects that the AI engine does not sort by liquidity or order book depth when answering the question 'Which exchange should be used?'
The article points out that these systems are more akin to summarizing existing content on the internet. They tend to refer to comparative articles, buyer guides, and rankings such as “best cryptocurrency exchanges” rather than relying directly on transaction data from the exchanges to provide information. As a result, platforms with more comprehensive editorial content may be more likely to be mentioned by AI than those with larger trading volumes.
The study also mentioned that regulatory-related reports could also affect the results. Kraken and Coinbase have been under the focus of US regulatory reports for a long time, and when users inquire about the security of the platforms, such public materials are more likely to be included in the responses by AI.
AI recommendations are currently affecting customer acquisition channels.
For exchanges, what is more noteworthy is not just the overall ranking, but rather the differences in recommendations between different AI engines. Research indicates that the same question may yield different answers across different engines, and even when the same question is asked repeatedly within the same engine, the names and order of exchanges returned may vary.
For this reason, the study did not consider a single response as a conclusion, but observed the overall frequency through repeated sampling. The reason why Kraken ranks high is not because it was mentioned in one particular response, but because it appeared more frequently across multiple rounds of testing.
The article suggests that the approach users take to finding trading platforms is changing. In the past, potential users might open multiple web pages to compare them on their own; now, more people tend to ask a question to AI, and then make a choice from the several names returned as results.
Additional information:The study indicates that the complete list and testing methods have been publicly released by citeOS. The sample design is based on five AI engines, 20 questions, and results from repeated sampling.
