Does Grok cite your website?

Updated 2026-06-13 · maintained by Unsourced

Grok has a reputation for pulling from X and talking with confidence, so it's easy to write it off as a citation channel. That's out of date. Grok now combines live web search with X and shows inline citations you can click. The real catch is different: xAI doesn't publish a documented crawler the way OpenAI and Anthropic do, which makes verifying Grok's footprint harder. That's exactly where evidence beats assumption.

What most guides get wrong

• Grok just makes things up from X. Not anymore. It retrieves from the live web and from X, and attaches citations that link to the sources it used, so it can cite your site directly.
• I can spot Grok in my logs. You mostly can't, cleanly. xAI doesn't publish the named crawler, IP ranges and reverse-DNS footprint that let you confirm a visit the way you can with OpenAI or Anthropic. Identity here is genuinely harder, which is why measuring the citation outcome matters more than watching for a crawler.
• A mention proves Grok read me today. With a thinner verification trail, the safest signal is the citation itself, tracked over time, rather than a log line you can't confirm.

How Grok surfaces sources

Grok answers by combining its model with real-time retrieval from the open web and from X, and presents citations that link out to the sources behind an answer. Because much of its edge is recency, fresh and clearly written pages tend to travel best. Standard, openly crawlable HTML is what any retrieval step can read.

The crawlers behind Grok

• No documented public crawler xAI does not publish a separately-named web crawler with verifiable IP ranges and reverse DNS, so confirming a request is genuinely Grok is not reliably possible today. Treat any “Grok” user-agent as unverified.

When you can't verify the visitor, you verify the result instead: track whether Grok actually cites you, rather than trusting a crawler claim you can't check.

How to check whether Grok cites you

You can put your audience's questions to Grok and watch for your name, but its answers swing between sessions and lean on whatever's fresh, so a single look isn't evidence, and with no documented crawler to confirm a visit, the citation itself is the signal that counts. Unsourced is built unlike tools that fire a generic prompt list and hope. We construct the test questions from your own footprint: the queries you already rank for in Google and Bing Search Console, the topics you publish on, and the pages real visitors land on, written the way someone would actually ask. We run them on a schedule, record each time Grok cites your domain and which competitors it picks instead, and, where Origin Signal can, show whether a mention came from a live pull or older data. Measurement grounded in your real demand, not a canned list.

How to improve your chances

• Be openly crawlable. Keep your content on the public web in server-rendered HTML, with nothing blocking general retrieval.
• Lead with freshness. Grok leans hard on recency, so current, clearly dated pages have an edge.
• Have a presence on X. Grok blends X with the web, so visibility there can feed what it surfaces.
• Write quotable answers. Direct, well-structured pages are easier to cite than sprawling ones.
• Measure the outcome. With no reliable crawler to verify, tracking the citations themselves is the honest signal.

Nobody can guarantee a Grok citation, and with limited crawler verification, anyone who promises one is guessing twice over. What you can do is stay openly crawlable, publish fresh answers worth citing, and measure the result honestly, which is the whole point of Unsourced.

Common questions about Grok

Other AI assistants