How Much Can We Trust LLM Search Agents? Measuring Endorsement Vulnerability to Web Content Manipulation

Abstract

Large language model (LLM)-based search agents synthesize open-web content into actionable recommendations on behalf of users, creating a risk that attacker-published pages are transformed into endorsed claims. We introduce SearchGEO, a controlled evaluation framework for measuring endorsement corruption in LLM-based web-search agents, combining a web-evidence manipulation pipeline, a five-mode attack taxonomy, and multiple output-level metrics. We evaluate 13 LLM backends on 308 cases each. Res

Categories

@misc{chen2026howb,
  title = {{How Much Can We Trust LLM Search Agents? Measuring Endorsement Vulnerability to Web Content Manipulation}},
  author = {Yimeng Chen and Zhe Ren and Firas Laakom and Yu Li and Dandan Guo and Jürgen Schmidhuber},
  year = {2026},
  month = jun,
  eprint = {2606.16821},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2606.16821}
}