The landscape of AI licensing is shifting from static, one-off training deals to more dynamic, usage-based models that respond to today’s need for real-time content. This new approach, often known as ‘grounding’, lets AI systems fetch current information as needed rather than relying on data that could be months old.
Publishers are now seeing the benefit of recurring revenue streams instead of a single lump sum payment. When fees are tied to how often content is accessed, every click counts. Sure, engineers might debate terms like ‘content inference compute’ or even ‘retrieval augmented generation’ (RAG), but for media houses, the takeaway is clear: ongoing income that scales with usage.
If you’ve ever wrestled with the challenges of monetising digital content, you’ll appreciate the nuance behind grounding deals. Major names like The New York Times and News Corp may still be riding big training deals, but for many, shifting to real-time content retrieval offers a more robust and fair model. This shift also underscores the importance of proper attribution—every time content is used, it’s appropriately credited and licensed.
Recent agreements, such as the one between Gannett and Perplexity, illustrate how these models are evolving. Instead of flat fees, publishers now engage in ad revenue sharing and pay-per-usage arrangements. This means whether it’s pay-per-query or pay-per-crawl, the deal is fine-tuned to our fast-paced media environment, ensuring that both content producers and AI platforms benefit.
Simply put, transitioning from static training models to grounding agreements enables publishers to better monetise their content while delivering up-to-date information to audiences. As AI tools become more integrated into our daily digital experience, these evolving deals provide a more sustainable, fair approach for everyone in the content ecosystem.
