DuckAssistBot: DuckDuckGo's AI answer fetcher, and how to opt out

Where DuckAssistBot fits

DuckAssistBot does one job: when DuckDuckGo produces an AI-assisted answer, it fetches pages in real time so the answer can cite them. DuckDuckGo says those answers prominently cite their sources and that the data is not used to train AI models.

Our directory files it under "User-triggered fetch", and that label needs a caveat. It fetches in real time because someone ran a search that produced an AI answer, which is why it sits alongside ChatGPT-User and Claude-User. But it is not a fetcher a person points at a specific URL, it obeys robots.txt with no user-initiated exemption, and it is not a general search indexer either. You could reasonably call it a search bot. What matters for your robots.txt is the same either way: it only affects AI-assisted answers.

What blocking DuckAssistBot does, and what it doesn't

Disallow it and your pages won't be used or cited as sources in DuckDuckGo's AI-assisted answers. DuckDuckGo says opting out does not affect organic search rankings or inclusion in results, so your normal DuckDuckGo listings stay as they are.

Because the data isn't used for training, blocking it doesn't buy you a training opt-out. If you want to be cited when someone asks DuckDuckGo a question about your category, leave it allowed.

How to block or allow DuckAssistBot

User-agent: DuckAssistBot
Disallow: /

DuckDuckGo says the crawler stops visiting 72 hours after you add the rule, so don't expect requests to stop the same day. To allow it explicitly, use Allow: / instead. A named group replaces your User-agent: * rules for DuckAssistBot, so a wildcard block will catch it unless you give it its own group. robots.txt applies per host, so each subdomain needs its own file. For questions, DuckDuckGo lists crawling@duckduckgo.com.

Checking that a visit is really DuckAssistBot

The documented user agent is short: DuckAssistBot/1.2; (+http://duckduckgo.com/duckassistbot.html), with no Mozilla prefix and an http link. Log filters that only look for browser-style bot strings can miss it. Since anyone can send the string, compare the source address with DuckDuckGo's published IP ranges.

The live check above sends a request with DuckAssistBot's token and compares it with a browser request, which shows whether a firewall or CDN is blocking it. It can't come from DuckDuckGo's IPs, so IP-keyed rules may still differ. The robots.txt checker shows whether a wildcard rule is catching DuckAssistBot.