Search papers, labs, and topics across Lattice.
This paper introduces a taxonomy of advertising styles for LLMs, considering explicitness and type of appeal, to simulate how advertisers might evade detection in RAG systems. They evaluate the robustness of various ad-detection approaches, including entity recognition-based models, under changes in advertising style. The results show that entity recognition-based models are effective and robust, while lightweight models like random forests and SVMs are brittle, suggesting a need for more efficient and robust ad-blocking techniques.
Entity recognition models can effectively spot RAG-powered native ads, even when advertisers try to disguise them with different styles.
Large language models (LLMs) enable a new form of advertising for retrieval-augmented generation (RAG) systems in which organic responses are blended with contextually relevant ads. The prospect of such"generated native ads"has sparked interest in whether they can be detected automatically. Existing datasets, however, do not reflect the diversity of advertising styles discussed in the marketing literature. In this paper, we (1) develop a taxonomy of advertising styles for LLMs, combining the style dimensions of explicitness and type of appeal, (2) simulate that advertisers may attempt to evade detection by changing their advertising style, and (3) evaluate a variety of ad-detection approaches with respect to their robustness under these changes. Expanding previous work on ad detection, we train models that use entity recognition to exactly locate an ad in an LLM response and find them to be both very effective at detecting responses with ads and largely robust to changes in the advertising style. Since ad blocking will be performed on low-resource end-user devices, we include lightweight models like random forests and SVMs in our evaluation. These models, however, are brittle under such changes, highlighting the need for further efficiency-oriented research for a practical approach to blocking of generated ads.
