How to get cited by Grok: a guide for B2B SaaS teams
Grok cites what its two search tools bring back. xAI's developer documentation describes a web search tool that searches the web and opens pages, and an X search tool that finds posts, accounts and whole threads on X, and a single answer can cite results from both. For a B2B SaaS team that means two routes into Grok's answers: the web pages that answer a buyer's question (comparison and pricing pages, public documentation, review-site listings and the roundups of your category), and your company's posts and profile on X, a route particular to Grok, since xAI documents a tool of its own for searching X. What xAI leaves out matters as much. As of 17 October 2026 its documentation has no page for a crawler, a user agent or robots.txt, so there is no documented Grok agent to allow or block by name. Serve your key facts as plain text in the HTML, make your X account state plainly what the product does, and judge progress by the sources Grok's answers actually carry.
The Rank Master measures Grok every month alongside ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews and Google AI Mode, for B2B SaaS companies, with published pricing and no annual lock-in. Book a 30-minute call: we will look at your category live and show you what the engines say when a buyer asks for a recommendation.
Every statement below about Grok's tools comes from xAI's developer documentation, read on 17 October 2026. Those pages describe Grok as developers use it through xAI's API. xAI publishes no matching account of how grok.com, the Grok apps or Grok inside X pick their sources, so treat the API as the closest documented view of the product rather than a full description of it.
What xAI documents about how Grok finds sources
When a developer hands Grok tools, xAI's tools overview describes a loop: the model works out what it needs, chooses between calling a tool and answering, runs the tool, reads what comes back and repeats until it has enough, then replies with citations where they apply. Two of the built-in tools search public sources, and they work quite differently.
| Web Search | X Search | |
|---|---|---|
| What it does | Searches the web in real time and browses pages to read them | Searches posts, user profiles and threads on X |
| How it searches | Search calls, one variant returning snippets, plus functions that open a page | Keyword queries in X's advanced search syntax, ordered by Top or Latest; meaning-based search; account lookup by name or profile description; a single post fetched with the posts above it and its replies |
| Images and video | Can inspect images on the pages it browses, if the developer turns this on | Can view images in posts and watch their videos, each off by default |
| Limits a developer can set | Search only within, or leave out, up to five domains | Include or exclude up to 20 handles; restrict posts to a date range |
| Cost to a developer | $5 per 1,000 calls | $5 per 1,000 posts and $10 per 1,000 profiles fetched |
Sources: xAI's Web Search, X Search, Tool Usage Details and pricing pages, read 17 October 2026.
Two details shape everything else on this page. First, the usage notes list the functions behind the web tool as web_search, web_search_with_snippets, browse_page, open_page and open_page_with_find. Grok is not limited to search snippets: the tool can open your page and read it, so what the page says in its text is what Grok has to work with. Second, X search is not a single keyword lookup. It can find an account by what its profile says and pull in a whole conversation, so your profile, your posts and the replies beneath them are all material.
xAI's release notes date both tools to October 2025, when its server-side tools became generally available. Developers can also narrow either tool, which is one reason the same question can come back with different sources in different products built on Grok.
Two lists of sources, and what each tells you
xAI's citations page separates two things. Every response carries a full list of the URLs the agent came across while researching, returned by default. xAI notes that some of those were looked at and set aside as not relevant enough, so being on that list does not mean being used. Separately, the answer text can carry inline citations: numbered links that start at 1 and reuse a number when the same source returns. xAI adds that citing is left to the model's judgement, so an answer can carry no inline citations at all even with the option on.
In xAI's Python SDK, the structured citation data also labels each inline citation as a web citation or an X citation. One answer can therefore point at your pricing page and at a post on X side by side.
For a SaaS team the split is a useful diagnosis. A page that appears in the full list but never inline was found, and possibly used, but not cited inline. One answer proves little, because the model can leave out inline citations altogether; if the pattern holds across repeated runs while other pages take the inline citations, those pages are probably answering the buyer's question more directly or more completely. A page that never appears at all points to an earlier problem: Grok's searches are not finding it, or its fetch is not getting through.
The gap: no documented crawler
OpenAI, Anthropic, Perplexity and Google each publish a page that names their agents and the robots.txt tokens that control them. xAI does not. On 17 October 2026 we checked xAI's documentation index (llms.txt), its sitemap of 236 pages and llms-full.txt, which joins every documentation page into one file: none names a crawler or a user agent, and none mentions robots.txt. That is a statement about the documentation, not about how Grok behaves. Grok's web tool does open pages, and those requests come from somewhere, but xAI does not say what they call themselves, which addresses they use or whether they read robots.txt.
Three things follow for a SaaS site.
- xAI documents no Grok line for robots.txt. A group for a guessed name is not something you can rely on. If Grok's fetcher reads robots.txt at all, xAI does not say which group it would follow.
- Know what your bot protection does with automated visitors it cannot identify. There is no documented Grok agent or IP list to exempt, so a blanket rule against unidentified automated traffic may turn Grok's fetches away along with everything else. Weigh that against the reason the rule exists; the aim is to know, not to switch protection off.
- Serve the facts in the HTML. xAI does not say whether its fetcher runs JavaScript. A pricing table or comparison grid that only appears once scripts run is a gamble with any agent that may not run them. Our guide to AI crawlers and robots.txt sets out what the other vendors document, agent by agent.
One more detail from the usage notes: xAI lists an attempt to browse a page that does not exist, or to fetch a deleted post on X, among the tool calls that fail, after which the agent tries other routes. Keep URLs stable, redirect the ones you retire, and think twice before deleting a post that states a fact you still want found.
Make the web pages do the work
Web search is one of Grok's two documented routes, and for buying questions it is the route Grok shares with every other engine. The pages that get a SaaS product cited elsewhere do the same job here: a comparison page with a dated feature and price table, an honest alternatives page, a pricing page that states plan limits, and public documentation for integrations and security. That ground is covered for all seven engines in our guide to getting a SaaS product recommended by AI engines, so it is not repeated here.
Make your presence on X citable
What follows is our reading of the X search documentation, not advice from xAI. The documentation says what the tool can do; it says nothing about how Grok weighs what it finds.
- Say what you are in the profile. X search can look up accounts by name or by description. A company profile whose name and bio state the category and the buyer, such as "payroll software for remote-first companies", gives that search something to match. A slogan gives it nothing.
- Put the facts in the words of the post. Keyword search runs on X's own advanced search operators and the semantic search matches meaning, but both work on what a post says. Launches, pricing changes, new integrations and security reports belong in plain words in the post, with the product's name, not only in a screenshot or a video. The image and video options exist, but both are off by default in the API.
- Write threads to be read as a whole. A thread fetch returns a post with the posts above it and the replies below it. A launch thread that carries the details in its first post, with replies that answer customers' questions accurately, reads well as one piece. A thread whose replies are full of unanswered complaints reads that way too.
- Correct where the old fact sits. When a price or a plan changes, post the change, and reply to the old post that carries the outdated figure. Because a thread fetch brings back the replies with the post, the correction travels with it.
None of this calls for posting more. A few clear, factual posts and an accurate profile give X search more to work with than a stream of promotion.
How to test what Grok cites
In the product. Put your buyers' questions to Grok on grok.com, in the apps and inside X, and record whether you are named and which links come back. Use the same discipline as for any engine: fixed wording, several runs, a date on every run. Our guide to measuring brand visibility in ChatGPT and AI answers sets out a prompt panel and a scoring sheet that work for Grok unchanged.
Through the API, for the full source list. A developer can send the same prompts through xAI's API with web search and X search switched on and keep both lists for every answer: the inline citations, and the full list of sources the agent encountered. The second list answers the question the first cannot, which is whether Grok ever saw your page. The cost is small at panel scale: $5 per 1,000 web search calls and $5 per 1,000 posts fetched from X, plus the model's tokens. Remember that the developer chooses the tools and settings, so API answers show how Grok behaves with those settings, which may not match the consumer app.
In your analytics. xAI does not document what referrer, if any, its links pass on. Check your referral data for visits from grok.com before building a report around them, and do not count visits from x.com as Grok traffic, since ordinary links shared on X arrive from there too.
Questions B2B SaaS teams ask about Grok citations
Does Grok have a crawler you can allow or block in robots.txt?
Not one that xAI documents. As of 17 October 2026 its documentation names no crawler or user agent and does not mention robots.txt. Grok's web search tool does open pages, but xAI does not say how those requests identify themselves or whether they read robots.txt.
Does Grok only cite posts on X?
No. xAI documents a web search tool alongside X search, and its Python SDK labels each inline citation as a web citation or an X citation.
Can Grok read a pricing page that loads with JavaScript?
xAI does not say whether its web tool runs JavaScript. Put prices, plan limits and comparison tables in the HTML your server sends, so that being read does not depend on it.
Is Grok through the API the same as Grok on grok.com or in X?
xAI documents the API's tools and settings. It does not publish how its consumer products configure them, so results from the API are the best documented guide to the consumer product, not a copy of it.
How this guide was researched
The facts about Grok come from xAI's documentation, all read on 17 October 2026: the tools overview, the Web Search, X Search, Citations and Tool Usage Details pages, the pricing page, the release notes, and the consumer Grok overview and FAQ. The absence of a crawler page was checked against the documentation index, the sitemap and the full-text file the same day. Where xAI's documentation is silent, on robots.txt, JavaScript, referrers and how sources are ranked, this guide says so rather than filling the gap with guesses.
Grok is one of the seven engines The Rank Master tracks every month for B2B SaaS companies (ChatGPT, Perplexity, Gemini, Claude, Grok, Google AI Overviews and Google AI Mode), reported in a live dashboard and a monthly report: who gets named for your buying questions and which sources the engines quote. Plans have published pricing, a 90-day initial term and then month-to-month on 30 days' notice, and your pages, briefs, data and contacts stay yours. Book a 30-minute call to see what the engines say about your category today and get a plain answer on whether we can help.