How to Get Cited by Grok: What xAI Documents and What It Does Not

Grok is one of the few AI assistants with two separate windows onto the world: the open web, and the live stream of posts on X. That makes "how do I get cited by Grok?" a slightly different question from the same question about ChatGPT or Perplexity, and it is one that a lot of guides answer with confident claims nobody can check.

This guide takes a stricter approach. xAI does not publish ranking factors for Grok, and we will not invent any. Instead, every recommendation below is labelled: Documented by xAI where xAI's own documentation says it, Prerequisite where it is simply necessary for Grok to read your page at all, and General GEO practice where it is good practice across AI assistants but not confirmed for Grok specifically. Start by running the AI crawler checker, because the first requirement is that Grok can reach you.

How Grok finds and cites sources

xAI's developer documentation describes the tools Grok uses to look things up. These are the API tools developers call, and the closest public description of how Grok gathers information.

• Documented by xAI Web Search "enables Grok to search the web in real-time and browse web pages to find information" (Web Search tool documentation).
• Documented by xAI X Search lets Grok "perform keyword search, semantic search, user search, and thread fetch on X" (X Search tool documentation).
• Documented by xAI Citations come in two forms. "All citations" lists every source the agent encountered; inline citations are links placed in the answer where the model references a source (Citations documentation).

Two details in the citations documentation matter more than anything else in this guide.

1. Documented by xAI "Not every URL in this list will necessarily be directly referenced in the final answer. The agent may examine a source during its research process and determine it is not sufficiently relevant."
2. Documented by xAI "Enabling inline citations does not guarantee that the model will cite sources on every answer. The model decides when and where to include citations."

In plain terms: Grok can read your page and still not cite it, and some answers cite nothing at all. Getting read is necessary. Being the clearest, most relevant source for the question is what turns a read into a citation, and xAI does not say how it measures that.

The two Groks you are optimising for

There is a detail here that most guides miss. "Grok" is both a consumer assistant, in the Grok apps and on X, and an API that developers build into their own products. The documentation above describes the API tools. When a company builds a support bot, a research tool or an internal assistant on Grok, it uses these same tools, and it gets to configure them.

Documented by xAI All five come from xAI's Web Search tool documentation and X Search tool documentation. The practical lesson is that the same page can be visible in the Grok app and invisible in a product built on Grok, because that product's developer chose its sources. That is outside your control, so the best response is to be a source a developer would choose: accurate, specific and clearly owned.

What we do not know

Being clear about the gaps protects you from wasted effort.

• xAI has not published which crawler fetches pages for Grok. The names in use, GrokBot and xAI-SearchBot, are reported by public registries, not xAI. Cloudflare's AI crawler reference lists no xAI crawler.
• xAI has not said which search index Web Search uses, or whether it runs its own.
• xAI has not published how it weighs sources: authority, freshness, X engagement or anything else.
• xAI has not said whether pages fetched for answers are also used for training.

Step 1: Make sure Grok can read your pages

Prerequisite Nothing else matters if the fetch fails. Three things commonly stop it.

robots.txt

Allow both reported names in one group, especially if your User-agent: * group is strict:

User-agent: GrokBot
User-agent: xAI-SearchBot
Allow: /

The robots.txt generator has an "Allow Grok" preset that writes this for you. Because xAI has not documented either name, this is the best available rule, not a guarantee. The full background is in the Grok crawler user agent guide.

Firewalls

Bot-management tools often challenge crawlers they do not recognise, and xAI's are not on Cloudflare's verified list. A challenge overrides any robots.txt allow, silently. Check your firewall events for xAI names, and see firewalls that block AI crawlers by accident.

JavaScript

Many AI fetchers do not run JavaScript. If your answer only appears after scripts load, assume Grok may not see it. Test a page with the AI crawler diagnostic, which checks the raw HTML along with five other gates.

Step 2: Answer the question on the page

Documented by xAI The citations documentation says the agent drops sources it judges "not sufficiently relevant". General GEO practice So the most useful thing you can do is make relevance obvious.

• Put the answer near the top. If the page is about your opening hours, state them in the first paragraph, in text.
• Use the words people ask in. A heading that reads like the question, followed by a direct answer, is easy for any system to match.
• Be specific. Prices, dimensions, dates, locations and named sources give an answer something concrete to quote.
• One topic per page. A page that answers one question well is easier to cite than a page that touches twenty.

None of this is unique to Grok. It is the same advice that helps with ChatGPT, Perplexity and Google's AI features, which is exactly why it is a safe bet.

Step 3: Keep it current and dated

Documented by xAI Grok's Web Search is described as real-time, and X Search reaches "real-time social media content". General GEO practice For questions where timing matters, such as prices, availability, news and product versions, a page that says when it was last updated, and is actually updated, gives an answer engine a reason to prefer it over an older one.

• Show a visible "last updated" date, and keep it honest: only change it when the content changes.
• Use dateModified in your structured data to match.
• Retire or redirect pages that are out of date, rather than leaving them to be quoted.

Step 4: Use X deliberately

Documented by xAI X Search gives Grok direct access to posts, users and threads on X. General GEO practice That makes your presence on X a second, separate route into Grok's answers, and one that robots.txt does not touch.

• Post the facts, not just links. A post that states your new price or launch date is itself searchable content. A bare link depends on the fetch succeeding.
• Keep your profile accurate. Your name, description and website link are what a user search returns.
• Correct errors publicly. If posts about your brand are wrong, a clear reply from your own account is part of the record Grok can find.
• Link X and your site both ways. It helps any system connect the two.

We do not know how Grok weighs X posts against web pages, or engagement against accuracy. So treat X as one more channel to keep accurate, not a ranking lever. And keep the link previews working: they come from Twitterbot, which is separate from Grok's crawlers.

Step 5: Make your pages easy to understand as data

General GEO practice Structured data does not guarantee citations in any AI assistant, but it removes ambiguity about what a page is and who stands behind it.

• Organization markup with your name, logo, website and social profiles, including your X account.
• Product, Service, LocalBusiness, Article or FAQPage markup as appropriate, with the same facts as the visible page.
• Clear authorship on articles: a named author with a profile page.

Documented by xAI xAI's Web Search has an option for "image understanding", which lets the agent "analyze images encountered during the search process". General GEO practice If key facts only appear inside images, such as a price list as a picture, put them in text as well. Use the schema generator for the markup.

Step 6: Be a source worth choosing

General GEO practice Every AI assistant has to decide which of several pages on the same topic to rely on, and none of them publish how. The signals below are the ones a careful human researcher would use, which is the most defensible guess at what an AI system approximates.

• Show who is behind the page. A named author with relevant experience, an About page and a real address or company details make a page easier to trust than an anonymous one.
• Cite your own sources. Link the study, the regulation or the manufacturer's specification you rely on. A page that shows its working is easier to check, and easier to cite.
• Say what you do not know. Pages that separate fact from opinion, and admit uncertainty, are less likely to be contradicted by other sources, and an answer engine comparing sources will notice contradictions.
• Be consistent everywhere. The same name, address, prices and claims on your site, your X profile, your Google Business Profile and directory listings. Conflicting facts give an AI no reason to pick yours.

This is slower work than editing robots.txt, and it pays off across every AI assistant and every search engine at once. It also lasts: a rumoured ranking trick stops working when the system changes, but an accurate, well-sourced, clearly owned page stays useful whatever the system does. For a broader checklist, see our AI SEO audit checklist.

Step 7: Measure, but carefully

There is no Grok equivalent of Search Console. What you can do:

• Ask fixed questions on a schedule. Keep a list of the questions your customers ask, ask Grok each one weekly, and record whether you are mentioned or cited. Our guide to Grok Bot workflows for SEO teams shows how to automate this.
• Watch referrals. Visits from Grok and X show up in analytics as referrals. Track them over time.
• Check your logs. The Grok log checker shows whether the reported xAI names are reaching you at all.

Expect noise. AI answers vary between runs, users and regions. Look for trends over weeks, not changes from one day to the next.

Three illustrative examples

These are illustrations of how the steps apply, not reported results.

A local bakery

Customers ask Grok "which bakeries near the station open before 7am?". The bakery's opening hours were in a photo of the shop sign on its homepage, and its X profile had not been updated since it changed hours. It put the hours in text on the homepage, added LocalBusiness markup with openingHours, updated its X bio and posted the new hours. Every one of those changes is also what a human visitor needed.

A B2B software company

Prospects ask "which tools integrate with our accounting system?". The company's integrations page loaded its list with JavaScript from an API, so the raw HTML contained only a heading. It server-rendered the list, gave each integration its own page with a clear one-paragraph description, and added a dated changelog. It also posted each new integration on X with the name in the text, not just a link.

A product review site

The site had blocked every AI crawler by default after a scraping incident, which also blocked GrokBot and xAI-SearchBot. It allowed the search and answer crawlers it wanted to be cited by, kept training crawlers blocked, and replaced the blanket firewall rule with rate limiting. See blocking AI training while allowing AI search for how to split the decision.

Grok compared with other AI assistants

Most of this guide applies to every AI assistant. The differences for Grok are about evidence and channels, not tactics.

The last row is the one to remember. No major AI company publishes how it chooses sources. The difference with Grok is only that even the first step, which crawler reads your page, is undocumented. That is why the access steps above are worth getting exactly right.

What not to do

• Do not write for a rumoured algorithm. There is no published Grok ranking factor to optimise for.
• Do not hide text for AI. Hidden instructions or keyword blocks aimed at AI assistants look like manipulation to people and search engines, and xAI treats web content as untrusted input.
• Do not flood X. Volume is not accuracy, and spam damages the brand that the posts are about.
• Do not block Twitterbot to "protect" content from Grok. It only breaks your link previews.

Questions we get about Grok visibility

Is Grokipedia a way to get cited?

xAI's own documentation uses grokipedia.com as its example of a domain to restrict search to, which shows that xAI treats it as a source. We have no documentation on how pages get into Grokipedia or how it is weighted in Grok's answers, so we would not plan a strategy around it. Make sure facts about your organisation there are accurate if you appear, as you would on any reference site.

Does paying for X Premium help?

xAI does not say that it does, and we found no documentation linking subscriptions to citations. Treat any claim otherwise as unverified.

How long does it take for changes to show up?

Unknown, and likely variable. Grok's search is described as real-time, but which index it searches and how often it is refreshed are not documented. Make the change, then check your fixed questions weekly.

Should I write content specifically for Grok?

No. Write for the people asking the question. Everything that helps Grok cite you helps every other AI assistant and every search engine, and nothing in xAI's documentation suggests a Grok-only format.

A summary table

Where to start if you have one hour

1. Fifteen minutes: run the free AI crawler check and fix any robots.txt rule that blocks GrokBot or xAI-SearchBot, if you want Grok to read you.
2. Fifteen minutes: run your three most important pages through the AI crawler diagnostic and note anything that is not in the raw HTML.
3. Fifteen minutes: ask Grok the five questions customers most often ask about your business, and write down what it says and which sources it cites.
4. Fifteen minutes: update your X profile and post your three most important facts in text.

Repeat step 3 every week with the same five questions. That list is your baseline, and it will tell you more about your Grok visibility than any checklist. Note the date, the exact wording of each question, and whether you were signed in, because all three can change the answer. After a month you will have a clear picture of which questions you are already winning and which ones need a better page.

Your checklist

• □Allow both reported xAI names in robots.txt, and check your firewall is not challenging them.
• □Confirm your key answers are in the raw HTML, using the AI crawler diagnostic.
• □Rewrite your most important pages so the answer comes first, in the words people ask.
• □Add an honest "last updated" date and matching dateModified.
• □Make sure your X profile is accurate and your key facts are posted in text.
• □Add Organization markup that links your site and your X account.
• □Start a weekly list of fixed questions and record whether Grok mentions or cites you.

For the full picture of how Grok reaches websites, read Grok Bot vs GrokBot. For AI assistants in general, see our guide to GEO vs SEO.