Fastly

Network TTFB: DNS, TLS, HTTP, CDN, Cache Rules

This technical guide explains how Time to First Byte (TTFB) is affected by network and delivery layers — DNS, TLS handshake, HTTP protocol and connection reuse, CDN routing/PoP selection, and cache rules — and provides diagnostic and retest recommendations for web teams. It emphasizes that theme or plugin changes should follow network-level fixes, shows how to interpret lab tools (PageSpeed Insights, Lighthouse, WebPageTest) to split TTFB into DNS/TCP/TLS/Waiting, and gives checklist actions (DNS TTLs/CNAMEs, enable TLS 1.3/OCSP stapling, enable appropriate CDN HTML caching or stale-while-revalidate, schedule synthetic monitoring) to reduce cold-start and geography-specific TTFB regressions.

Publishers Use LLM Honeypots to Fight AI Scrapers

Publishers and e-commerce brands are experimenting with “LLM honeypotting,” a deception technique designed to waste the compute and pollute the models of large-scale web scrapers and LLM builders. Approaches include proof-of-work challenges, creating endless plausible-but-useless content mazes, and feeding statistically coherent nonsense to degrade scraped data. The tactic is early and bespoke: some industry figures argue it can alter the economics of scraping, while critics say it’s easy to detect, costly to operate, and could have negative effects on the open web. Implementation decisions depend on site complexity, cost trade-offs, and platform infrastructure (CDNs/edge platforms) that can reduce the publisher’s incremental compute burden.

Publishers launch SPUR content telemetry standard

Publishers have formed the Standards for Publisher Usage Rights (SPUR), a publisher‑run coalition to build a content telemetry standard that tracks how AI systems access and use journalistic content. SPUR’s founding members include major news organisations such as the BBC, Financial Times, The Guardian, Sky, The Times of London, MediaHaus and, this week, the Associated Press. A technical working group published the telemetry standard on June 12 and opened it for public comment until July 24; the proposal defines five event types (retrieved, grounded, cited, displayed, engaged) and a data schema for reporting usage back to publishers. Microsoft, CDN vendor Fastly and several licensing/infra startups have provided feedback or signalled implementation interest. The coalition aims to move AI usage from opaque scraping toward usage‑based licensing publishers control.