Taboola Deploys LLM-Powered Argus Ad-Review System

Global advertising technology company Taboola has long faced slow and inconsistent manual ad reviews. As digital marketing increasingly demands speed and compliance, accelerating approvals while maintaining content quality has become an operational priority. The company has therefore turned to generative AI and automation to overcome the limitations of conventional human screening, improve ad-delivery efficiency and strengthen trust in its platform.

Taboola announced in July 2026 that it had used the Harness engineering architecture to develop Argus, an ad-review system powered by a large language model (LLM). The multimodal system can interpret text and images simultaneously. Since its deployment, it has cut review times to minutes and reduced manual workloads by 75%. Its architecture comprises five principal layers: control flow, context, tools, memory and CI/CD automation.