Did the Czech central bank, or CNB, announce at the start of 2025 that it would utilize ChatGPT to forecast inflation rate developments? No, that's not true: The bank did not announce plans to use ChatGPT to forecast inflation. Instead, in early 2025, the CNB published a blog post describing an experiment it conducted testing AI's initial potential for inflation forecasting. The study found that while artificial intelligence sometimes matched or outperformed analysts, its results were also unreliable due to methodological issues.
The claim appeared in a video (archived here) published on TikTok on June 1, 2025. The person in the video says in Czech, translated by Lead Stories staff:
So imagine, the Czech National Bank is reportedly going to forecast inflation using ChatGPT. I think it is a genius idea, perhaps they should also invite Mr. Cibulka with a dowsing pendulum or the guy from Dobris with a rod, he is pretty good too.
This is what the post looked like on TikTok at the time of writing:
(Source: TikTok screenshot taken on Wed Jun 4 09:16:04 2025 UTC)
Petr Cibulka (archived here), whose name is mentioned in the quote from the TikTok video, is a Czech political activist who uses a pendulum to predict major future events and often disseminates conspiracy theories. It is unclear who the person in the video refers to as "the guy with a rod from Dobris." Dobris is a town in the central part of the Czech Republic, and water dowsing is the practice of trying to locate underground water sources using a rod, unsupported by science.
The Czech central bank (archived here) has not announced plans to use AI for its macroeconomic forecasts, CNB spokeswoman Petra Vlčková (archived here) confirmed in an email to Lead Stories on June 10, 2025. The bank currently relies on its g3+ (archived here) model for forecasts, which does not incorporate any AI features.
The statement in the above video is based on an experiment with AI that CNB conducted and whose results were published on its blog on February 4, 2025 (archived here). The blog explains it was an initial test to see whether and how AI could contribute to macroeconomic forecasts. While the test found that the AI-generated results matched or even "outperformed" professional analysts and even the CNB's model, in some cases, for certain periods, the blog also cautioned that there were methodological limitations to using AI.
The blog post explains:
The main challenge remains its black-box nature, making it difficult to determine which factors influenced the predictions. Furthermore, these models exhibit limited reliability when applied to data outside their training set. The forecasts also contain an element of randomness, making them difficult to replicate.
According to Vlčková, CNB experts, after the first-ever revision of the bank's analytical framework, are currently developing two additional alternative forecasting models to complement existing tools. While some AI use in forecasting is possible in the future, it would only serve as one input among many, she stated.
The CNB's stance aligns with the 2024 report (archived here) from the Bank for International Settlements (archived here) regarding the use of AI by central banks. While the report encourages the monetary authorities to embrace the benefits of AI, it also warns against risks such as data bias, lack of transparency, cybersecurity threats, and challenges in explaining AI-driven decisions. According to the report, AI should complement human judgment, with strong governance and oversight to mitigate these risks.
