Content protection
Your content trains their models. You get nothing.
AI companies scrape your content to train models. Centinel catches them in real time — and lets you block, throttle, or license each one.
32%
of AI scrapes bypass robots.txt entirely
500K:1
crawl-to-referral ratio for AI bots
$0
is what most publishers get paid
How Centinel solves this
Detect
Identify which crawlers are accessing your content, how often, and what they're targeting.
Decide
Block, allow, challenge, or monetize each crawler individually from your dashboard.
Protect
Edge-level enforcement in under 2ms. Real users pass through untouched.
Frequently asked questions
- What constitutes content theft by an AI crawler?
- Content theft is any unlicensed extraction of your pages, paywalled or not, into a model's training or inference corpus. The crawler never cited you, never paid you, and the resulting model answers the same questions your article does. Tollbit reports that 30% of AI bot scrapes in Q4 2025 ignored explicit robots.txt directives entirely, which is the behavior in question.
- Does blocking GPTBot alone stop content theft?
- No. GPTBot is one named crawler out of 1,600 we track. Anthropic's crawl-to-referral ratio reached 500,000 to 1 per Cloudflare, and a long tail of retrieval crawlers, scraping-as-a-service providers, and training pipelines use generic user agents or residential proxies. A robots.txt line against GPTBot blocks the honest case and nothing else. Edge-level detection closes the rest.
- How fast does Centinel detect a new AI scraper?
- New crawler signatures land in our database within hours of first observation across our customer network. Per-customer, a previously unseen fingerprint hitting your site flags on the next request after it crosses our detection threshold. Most blocking decisions apply at the edge in under 2ms. You do not wait for a quarterly signature update.
- Can Centinel recover revenue from already-stolen content?
- Not retroactively. Models trained on past scrapes will not unlearn your content. What Centinel does from day one: stop ongoing extraction, flag which AI companies are still hitting you, and open a licensing path for the ones willing to pay. Several publishers on Centinel have converted active scrapers into paying license holders instead of chasing takedowns.
- How is this different from robots.txt?
- robots.txt is a request. Centinel is an enforcement layer. Tollbit's Q4 2025 data puts robots.txt bypass at 30% across AI bot traffic; OpenAI's ChatGPT-User agent bypasses at 42% of sites that explicitly block it. Centinel identifies crawlers by TLS fingerprint, HTTP/2 parameters, and 1,600 behavioral signatures, then blocks the request at the edge regardless of what the crawler claims.
Pick the next step that fits where you are
Demo, self-serve check, pricing, or a quiet email. Whichever maps to your stage.