---
id: "2026-10-15-cloudflare-openai-discovery-pilot-results"
url: "https://postcutoff.com/e/2026-10-15-cloudflare-openai-discovery-pilot-results/"
as_of: "2026-10-16T12:39:00+02:00"
date: "2026-10-15"
date_precision: day
category: business
importance: 2
confidence: high
status: [Confirmed]
sources: 3
editor: human editor
human_review: null
version: "2026-10-16"
---

As of: 2026-10-16 12:39 CEST.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/
Canonical page: https://postcutoff.com/e/2026-10-15-cloudflare-openai-discovery-pilot-results/

# Cloudflare–OpenAI discovery pilot results

Full title: Cloudflare and OpenAI report early results of their web-discovery pilot: network signals helped ChatGPT Search find hundreds of millions of new pages

In an Oct 15, 2026 post co-written by Cloudflare and OpenAI staff, the companies reported early results from a research pilot started in July 2026: signals from about 1% of Cloudflare's network (dead links, sites redirecting everything to the homepage, duplicates, change signals, page-level harm flags such as planted fake login forms) helped OpenAI discover and index hundreds of millions of new pages while cutting wasted crawler load. The post says the metadata is used for search and discovery, not for training models.

## Key facts
- Pilot announced July 8, 2026 (Business Wire); results post Oct 15, bylined by Jack Galilee (Cloudflare) and Henry Aspegren and James Balfour (OpenAI)
- Cohort: about 1% of Cloudflare's network; outcome: OpenAI discovered and indexed 'hundreds of millions' of new pages; no crawl-reduction figures given
- Signals: filtering of non-existent-file requests, catch-all redirects and duplicates; change signals for when to revisit; page-level harm flags
- Applies to sites 'that choose to make their content available to AI systems'; a signal does not grant access permission; Cloudflare's AI Crawl Control still governs access
- July quotes: Matthew Prince: 'By sharing our sophisticated network signals, we can find a better way to make AI search more efficient'; OpenAI's Nick Ryder: the pilot explores 'whether network-level insights can help us discover content more efficiently'

## What happened

Cloudflare passes metadata it sees as a network operator to OpenAI's search crawler. This lets OpenAI skip junk URLs, revisit pages
when they change and avoid individual malicious pages. The pilot covers about 1% of Cloudflare's network. The July 8 announcement
was missed at the time and is summarised here.

## Why it matters

Cloudflare spent the previous year building tools to block or charge AI crawlers. It is now helping one AI company crawl more
efficiently, on the condition that sites choose to let AI systems in. Discovery and freshness give AI search an edge, and a
network operator with this view of the web is a partner other AI search engines may also want.

## Your AI and this story

- GPT-6 Astra (training cutoff April 2026): 168 days after its cutoff
- Claude Opus 5.5 (training cutoff June 2026): 107 days after its cutoff
- Gemini 4 Argon (training cutoff March 2026): 198 days after its cutoff
- Grok 4.7 (training cutoff May 2026): 137 days after its cutoff

## Sources

1. [Cloudflare blog: Cloudflare and OpenAI: Making AI discovery better for everyone](https://blog.cloudflare.com/cloudflare-openai-discovery-agentic-web/) (blog.cloudflare.com, official)
2. [Barchart (Business Wire): Cloudflare announces research pilot with OpenAI (July 8)](https://www.barchart.com/story/news/3182635/cloudflare-announces-research-pilot-with-openai) (barchart.com, official)
3. [The Next Web: Cloudflare spent a year blocking AI crawlers. Now it's helping OpenAI index the web](https://thenextweb.com/news/cloudflare-openai-research-pilot-ai-search) (thenextweb.com, press)

## Changes

- 2026-10-16 (filed): Created