AI may fatally wound web's ad model, warns Tim Berners-Lee (www.ft.com)

Tim Berners‑Lee warned that large AI systems threaten the economic backbone of the open web: ad‑supported publishing. He argues that when LLMs and search agents ingest, summarize and re‑present web content directly to users, they sidestep pageviews, impressions and click‑throughs that fund journalism and free services. Because models are trained on scraped content and often deliver condensed answers (via retrieval‑augmented generation, embeddings and summarization), publishers can lose traffic and ad revenue while platforms or model providers capture the value. For AI/ML practitioners this is both a technical and policy problem. Technically, generative models change distribution — not just discovery — of content, undermining impression‑based monetization and analytics. Solutions being discussed include provable provenance and watermarking of model outputs, explicit content licensing or paywalled APIs, canonical metadata and publisher APIs (or decentralized protocols like Solid), and micropayment/token schemes to compensate creators. The warning underscores the need for engineering controls (rate limits, ingestion agreements, structured APIs), legal frameworks for dataset use, and standards for attribution so models can respect publishers’ rights without collapsing the web’s incentive structure.