AdKnob controls ad prominence and labels ads precisely in AI chats

AdKnob: Ad Intensity Control and Labeling for LLM-Native Advertising

Artificial IntelligenceComputers and SocietyMachine Learning

Summary

When chatbots powered by large language models show ads in their answers, it can be tricky to control how obvious those ads are and to clearly mark which parts are ads. The authors made AdKnob, a system that lets these models dial the amount of advertising up or down and label the ad parts exactly. They created new special words to help the model control ad levels and used a new way to trace ad content without slowing things down. Tests with several language models showed AdKnob works better and faster than past methods.

What this means in practice

  • For llm developers: Integrate controlled native advertisements in LLM-generated text with precise ad labeling and adjustable ad prominence.$Commercial implications: Enables LLM service providers to monetize conversational AI with configurable ad display while maintaining transparency.
  • For advertising technology teams: Automatically measure and label ad segments in AI-generated text to improve ad reporting and compliance without extra computational cost.

Authors

Woo Jae Kim, Seongho Keum, Joonsung Jeon, Suhyeon Ha, Sooel Son, Sung-eui Yoon

Abstract

Large language model (LLM)-native advertising has emerged as a viable monetization channel as LLM service providers seek sustainable revenue streams. However, integrating native ads into LLM responses introduces two largely understudied challenges: (1) controlling ad intensity -- how prominently advertising content appears -- and (2) ad labeling -- identifying which segments of the generated text constitute advertising. We propose AdKnob, a framework jointly addressing both challenges. For ad intensity control, we introduce new ad control tokens without pretrained semantic associations and align the model via direct preference optimization to generate ads at each desired level. For ad labeling, we propose an attention rollout-based attribution mechanism that identifies ad segments by tracing their attribution to input ad-related tokens, requiring no additional inference cost. We further construct MI-KnobSet, the first multi-level ad intensity preference dataset for LLM-native ads. Experiments across six LLMs and human evaluations show that AdKnob achieves uniformly spaced ad intensities where baselines collapse to a narrow range, achieves the highest ad labeling accuracy while being over 500x faster than the strongest baseline, and is preferred over baselines by human evaluators up to 89.6% of the time.