Fetcher bots, which retrieve content material in actual time when customers make queries to AI assistants, present completely different focus patterns. OpenAI’s ChatGPT and associated bots generated 68% of fetcher bot requests. In some instances, fetcher bot request volumes exceeded 39,000 requests per minute to particular person websites.
AI brokers verify a number of web sites when processing queries, producing greater visitors volumes per interplay than browser-based customers. That visitors runs by Fastly’s community for patrons utilizing the platform for content material supply.
Past visitors quantity, Fastly is seeing AI workloads working straight on its edge compute platform. Compton cited examples together with storage of enormous coaching datasets and prospects utilizing edge compute for inference and different AI-related processing duties.
Buyer technique shifts from blocking to optimization
The strategy massive media corporations are taking towards AI agent visitors has advanced over the previous six months. Compton stated the dialogue with media prospects has shifted from “‘how do you block it?’ to a way more nuanced and complex dialog now, about ‘how do you optimize for it?’”
Media corporations need their content material to stay related to AI fashions and chatbots however want instruments to handle how that entry occurs and implement current content material licensing agreements. Fastly developed AI bot mitigation expertise to deal with this requirement, permitting prospects to allow helpful bots whereas blocking dangerous ones.
The corporate additionally grew to become one of many first edge suppliers to help the Actually Easy Licensing (RSL) protocol. Fastly is working with massive media prospects as design companions to refine how the protocol will get applied on the edge. “It was an industry-developed protocol to primarily implement content material rights agreements associated to AI fashions,” Compton stated.
