OpenAI Taps Cloudflare Data for ChatGPT

On July 8, 2026, Cloudflare announced a research pilot with OpenAI. The company will share its network data to train ChatGPT and improve indexing accuracy.

A year against crawlers — and a sharp reversal

For the past twelve months, Cloudflare's stance was unambiguous: it built tools that let site owners block AI crawlers, the automated bots AI labs use to collect training data and index content.

That stance struck a chord. Publishers, bloggers, and web developers had complained about aggressive scraping by OpenAI, Google, Anthropic, and other AI companies, which added load to their servers and happened without explicit consent from content creators.

Now Cloudflare is changing roles, at least on an experimental basis. Under the new pilot, the company will test how its network data can help OpenAI crawl and index the web more accurately. Its site-protection tools aren't going anywhere: this is a research collaboration, not a change in business model.

Why Cloudflare's network data matters for AI search

Cloudflare handles traffic for more than a fifth of the entire internet. That gives it an exceptional vantage point. Instead of seeing the web from the outside, like a typical crawler moving from page to page, it sees it from the inside, through the stream of real requests to hundreds of millions of sites and services worldwide.

That data can give an AI search engine information that conventional crawling can't reach: which sites are genuinely alive and actively visited, how reliably a given site runs, and what content people are requesting in real time. It could offer a more accurate and up-to-date picture of the web than traditional search bots collect.

• Cloudflare's infrastructure reach: more than 20% of global web traffic
• Pilot goal: test whether network data improves the accuracy of OpenAI's AI search
• Partnership format: a research pilot, not a commercial product
• Announcement date: July 8, 2026

Context: the tension between AI and the open web

How AI companies harvest data from the open web has become one of the industry's biggest fights over the past two years. AI crawlers strain servers, consume bandwidth without any monetization for publishers, and raise copyright questions. Some publishers have put content behind paywalls specifically to keep AI scrapers out.

Cloudflare sits at the center of this debate. As an infrastructure provider, it serves both AI companies and the millions of websites trying to protect themselves from them. The crawler-blocking tools Cloudflare launched last year were an attempt to side with publishers, and they earned the company a reputation as a neutral defender of the open web.

The OpenAI pilot shows there's no simple choice between "for" and "against." Cloudflare is looking for a more nuanced position: rather than blocking AI systems wholesale, it wants a hand in shaping how those systems interact with the web.

What it means

If the research pilot shows that Cloudflare's data improves AI search, it could reshape the architecture of next-generation search engines, replacing mass page crawling with partnerships with infrastructure providers. For Cloudflare, that opens up a potential new revenue stream and a new role in the AI ecosystem.

Why does OpenAI need Cloudflare's data?

Cloudflare is testing whether its network data can improve the accuracy of ChatGPT's AI search.

Why did Cloudflare change its stance on AI crawlers?

Throughout last year, Cloudflare helped websites block AI crawlers, but it is now running a pilot with OpenAI to share its data.