The development of algorithms and the spread of social networking services cause filter bubbles that can contribute to the polarization of opinions and the breeding ground for fake news. Based on existing research that suggests virtual experiences influence people’s behavior, we hypothesize that experience able pseudo-filter bubbles can make people aware of the danger of filter bubbles and prevent them from falling into them. In order to generate experience able pseudo-filter bubbles, we aim to generate LLM agents with political ideologies that can imitate Internet actions, such as generating search queries. Furthermore, we mitigate the inherent left-leaning bias of the LLM so that we can generate LLM agents with more various ideologies. As a result, in addition to mitigating agents’ bias through prompt engineering alone, we have succeeded in generating agents with various ideologies, including extreme left-wing and almost extreme right-wing ideologies. This result expands the applicability of prompt engineering and can be the first step in creating pseudo-filter bubbles.