This talk traces the true sustainability cost of today’s generative-AI boom—from the “Ghibli effect” and cartoon-box trends to daily ChatGPT query volumes—contrasting GenAI’s resource footprint with traditional ML and the industry’s “scale at all costs” response.

We’ll examine how AI is flooding marketing and search: AI Overviews on billions of queries, query fan-out, bot/“dead internet” dynamics, and the waste created by the new vibe-coding workflow where endless prompting erodes trust and bloats emissions. We’ll map the downstream effects on our roles, ethics, and the reliability of the web, then pivot to pragmatic ways to cut impact without losing capability.

You’ll leave with concrete alternatives:
* when to do the work yourself,
* how to plan prompts and model choices,
* when to favor small/custom models or fine-tunes, and
* how to apply efficient architectures, pruning/quantization, and pre-trained APIs to replace default LLM-everywhere habits—complete with workflows, roadmaps, and resources.

This talk is about building systems that are mindful not only about the results needed and expected by us, marketers, to stay competitive in the digital landscape; but also of the environment and it's needs.

In this talk, Gijs will discuss what website structures AI models favor most when citing different sources in AI responses. Understanding how to correctly set your website up can be the difference of winning citations and mentions, or missing out on them completely.