Text this: Measuring susceptibility: A benchmark for conspiracy theory adherence in large language models