Cloudflare Says Most Web Traffic Is No Longer Human

Cloudflare: Fewer Than Half of Page Requests Now Come From Humans

Why it matters for your score

If most of the traffic touching a typical site is already non-human, AI agents are not a fringe case, they are close to being the majority visitor. Cloudflare's new Agent Readiness Diagnostics and Answer Engine Optimization scoring give site owners a way to measure whether they are being read, understood, and recommended by assistants like Claude and GPT, not just crawled and ignored.

Cloudflare Logs 206 Million Bot-Trust Events in a Single Day

Why it matters for your score

In one 24-hour window, Cloudflare recorded 206 million bot-evaluation events across more than 73,000 zones and found that traffic frequently shifts between human and agentic behavior mid-session. Static allow and block rules built around a fixed list of user agents will not keep up with that; expect crawler access decisions to move toward continuous, behavior-based trust scores instead of one-time robots.txt entries.

Perplexity Retires Sonar for a Single Agent API

Why it matters for your score

Perplexity is folding web search, page fetching, code execution, and MCP connections into one endpoint, with Sonar shutting down on September 27, 2026. If your team monitors AI visibility or builds anything on top of Perplexity's API, plan the migration now, since tools that still call Sonar after that date will simply stop working.

Anthropic Begins Watermarking Claude's Text Output

Why it matters for your score

Built on Google DeepMind's SynthID-Text method, the watermark is invisible to readers but detectable computationally, and Anthropic says it changes neither cost nor output quality. It is a preview of where content provenance rules are headed under the EU AI Act; expect disclosure and watermarking questions to become routine for any business publishing AI-assisted content, not just for the model providers.

Machine traffic has quietly become the majority of requests on many sites, turning crawler policy into general traffic management rather than a niche robots.txt setting.

Bot detection is moving from static allow-lists to continuous, behavior-based trust scoring, so a one-time crawler audit no longer holds up for long.

AI providers keep collapsing fragmented API tiers into single agent endpoints, a sign the infrastructure underneath AI search is still being rebuilt every few months.

Content provenance and watermarking are becoming default behavior at major AI labs, ahead of being a compliance checkbox that businesses have to answer for themselves.