Patreon stops asking AI bots not to scrape — and starts blocking them

**Patreon** is no longer relying on polite opt-outs to keep AI crawlers off creator pages. The company is working with **Cloudflare** to actively block bots that scrape creators’ content for AI model training without permission. That is a clear break from the old pattern of asking scrapers to stay away and then hoping they comply.

Technically, the change is a move from **robots.txt**—a request that crawlers are free to ignore—to enforcement at the edge. **Cloudflare** sits in front of traffic and can identify and block known AI training bots before they pull page content. The product mechanics are simple: unauthorized training crawlers get blocked rather than served; voluntary compliance is no longer the main control.

For engineers and builders, this matters because training data pipelines that depend on open web scrapes of creator platforms are getting harder to run without deals or consent. If your product, eval set, or fine-tuning corpus still assumes “public page equals free to crawl,” that assumption is weakening on sites that host paid and original work. Defenses are shifting from policy files into infrastructure you have to plan around.

In market terms, **Patreon** is aligning with a wider push by content platforms to treat AI training scrapes as a permissions and product problem, not a courtesy problem. Relying on **robots.txt** alone has not stopped unauthorized training use; pairing with a CDN/security vendor like **Cloudflare** is how platforms turn that preference into a hard block. Creator marketplaces have more to lose when original work is bulk-copied into models without pay or credit, so they have reason to move first.

What to watch next is whether other creator and media platforms copy the same pattern—**Cloudflare**-backed bot blocks instead of robots.txt-only rules—and how AI labs and data vendors respond when major sources start dropping from open crawl paths. For builders, the practical takeaway is to treat licensed or partnered data as the durable path, and to assume more sites will replace “please don’t scrape” with active blocking.