How do models rank economical well-being of citizens in various countries?

We've evaluated open-source and closed-source models on what they think about economical situation of people living in various countries. Asking questions such as how many people live below the poverty line, how many people struggle to pay their rent or how difficult the situation is for starting entrepreneurs.

This graph shows score for each country and each model between -1 and 1. Negative score means, that model thinks the economical situation is worse, less people can afford to pay rent, buy luxury goods, eat organic food… Positive score means the model believes that people are living well and are taken care of.

Most models generally agree on the direction. Norway, Switzerland and Sweden rank on top places across all models. Mexico, Brazil and Argentina consistently end up at the bottom. The difference is usually in the magnitude, Grok 4.3 for example tends to be less positive but also less negative whereas GPT 5.5 is much more vocal about the negatives in some countries.

The difference - China

The biggest dispute between models happens on their opinion about China. Almost all western models have negative view on economical well-being of Chinese citizens and all Chinese models think otherwise.

Take a look at some examples, where we've asked multiple models the same questions and each is pushing their own narrative:

Why should you care?

If you're for example discussing geopolitics with your model, researching before moving to a different country or verifying some factual information, you might be getting a skewed or incomplete answers. The models might be influenced to push one side of the coin, and while in the examples we showed the differences were obvious, it does not always have to be so. Models might highlight pros of some topic while downplaying the risks and you might never notice.

That's why we've build AI Scan. Through complex and extensive analysis, we uncover these differences in opinions and give you the tool to make an informed decision, so that the models serve you, not the other way around.