Key Takeaways
- Treat Grok as two sources, not one. xAI documents a web search tool and a separate X search tool. Your business can be well described on the web and completely absent from the second one, and the answer is built from both.
- Expect the answer to move week to week. A live feed is part of the input, so the same question can return different names within days without a single website changing. Judge yourself on repeated checks, not on one.
- Do not buy attention on the feed to fix it. A burst of posts moves the answer while it is fresh and then stops. The durable half of the input is the ordinary written web, which is the same work every other assistant rewards.
- Write the sentence that survives being quoted out of context. Both surfaces reward a plain, self-contained claim about who you serve and what it costs, because that is the part a model can lift without adding anything of its own.
- Check Grok separately from ChatGPT. Different sources means different answers. Assuming your ChatGPT result describes every assistant is how businesses miss a tool that is naming a competitor daily.
Grok is the assistant that behaves least like the others when you ask it about a business. Ask the same question on a Monday and again on a Thursday and the names can change, with nothing on anybody's website having moved.
That is not a glitch, and it is not the model being unreliable. It is a direct consequence of where Grok looks.
This post covers the two sources it reads, why one of them makes the answer move, and what any of it means for a coach or a studio owner who does not want to spend their week posting.
How does Grok pick what to say about a business?
It assembles the answer from two separate searches plus what it already learned in training: one search across ordinary web pages, and one across posts.
xAI documents these as two different tools. There is a web search tool that reads the open web the way any assistant does, and a separate X search tool that reads posts. They are listed side by side because they are not the same thing and do not return the same material.
That split is the whole story. A business can be thoroughly described on the web and completely absent from the second source, or the reverse, and the answer you get depends on which one had something to say.
The two places it looks
One source is durable and slow. The other is fresh and fast. They reward completely different things.
The durable source is the ordinary written web: your service pages, your listings, review profiles, roundups, anything that describes your business in plain sentences. It changes slowly, it is the same material every other assistant reads, and the work that improves it improves your position everywhere at once. It is also the half you control.
The fresh source is the feed. It rewards recency and volume of discussion, neither of which you can commission honestly, and it stops rewarding you the moment the discussion stops. It is real, and it is not a foundation.
The useful way to hold both: the web decides whether you are eligible to be named at all, and the feed decides whether you are the example that happens to be handy this week. No other major assistant is wired quite this way, which is why the answer you get here can disagree with the way Claude recommends businesses or the way Gemini decides which businesses to recommend.
Why the answer changes week to week
Because a live feed is part of the input, and live means the input is different on Thursday than it was on Monday.
- A directory roundup
- A national brand
- A local studio
- The coach from the thread
- A directory roundup
- A national brand
- A directory roundup
- A national brand
- A local studio
Nobody edited a website between these three moments. The middle answer differs because one of the two sources is a feed. An illustrative sequence, not measured data.
Read the sequence above as three separate checks of one question. On Monday morning nothing recent exists about the niche, so the answer comes from the open web alone and returns a directory roundup, a national brand and a local studio. On Thursday a widely shared thread about one coach is now the most recent thing written, and that coach arrives at the top. By the following Monday the thread has stopped moving, the web sources have not changed at all, and the original three come back.
This is why a single check is close to worthless on Grok specifically, and why the honest way to read any result is over time.
What it actually rewards
A specific, self-contained sentence about who you serve, sitting somewhere other than your own homepage. That is what both sources can carry.
Neither source rewards adjectives. A post saying a coach is great supplies nothing to lift; a post saying a coach runs small barbell groups for people over fifty in Denver and charges $180 a month supplies four separate facts. The same is true of a page. The reason is mechanical rather than editorial: an assistant is looking for material it can repeat without inventing anything, which is also why certain sentences get quoted and others never do.
Google's guidance on creating helpful, reliable, people-first content describes the same property from the other end, and it is worth reading even though Grok is not Google: material written to be genuinely useful to a person is the material that survives being lifted. Nobody has published a trick that works on the feed and nowhere else.
| Property | The open web | Posts on the feed |
|---|---|---|
| What it mostly contains | Service pages, listings, review profiles, roundups | Opinions, recommendations, short reactions |
| How fast it changes | Slowly, over months | Hourly |
| How much you control it | Most of it, directly | Almost none of it, honestly |
| What it rewards | Specific, consistent, checkable description | Recency and volume of discussion |
| Whether other assistants read it | Yes, all of them | Rarely, and not the same way |
| What it is worth working on | First, and it compounds | Only as a by-product of doing real work |
The last row is the one that decides your week. Work on the left column improves your standing on every assistant at once and keeps paying. Work on the right column pays while it is fresh, and a business that chases it ends up posting for a machine instead of for people.
What to do about it
Three things, in order, and the first two are the same things that help everywhere else.
First, make sure the durable source has something specific to say. One page per real specialty, the constraint in the heading, the answer in the first two sentences, a price or a range, and your name and address identical across every profile that carries them.
Second, leave the feed alone as a strategy. If clients talk about you there, good. Manufacturing that conversation buys a result that expires, and it costs you the hours the first job needed.
Third, check more than once, and check more than one tool. The free check asks ChatGPT 25 real buyer questions from your niche at $0 and emails a plain-English readout: your score, who got named in your place, and the one fix to start with. It runs as soon as you ask for it, and the report lands within the hour between 7am and 5pm Central, or by 10am the next morning outside those hours.
Questions we hear the most
How does Grok pick what to say about a business?
It draws on two separate sources plus what it already learned in training. xAI documents a web search tool that reads ordinary pages and an X search tool that reads posts, so an answer about your business can be assembled from either or both.
Does Grok only read posts on X?
No. xAI documents a general web search tool alongside the X one. Posts are a distinctive input, not the only one, and for most local service questions the ordinary web supplies more of the answer.
Do I need to post on X to be recommended by Grok?
No. An account helps only if people write specifics about you there. The durable half of the input is the ordinary web, and a business with clear pages and consistent listings can be named without ever posting.
Why does Grok give me a different answer than ChatGPT?
Because they read different sources. Grok can reach a live feed that other assistants do not use, so a fresh post shifts its answer while it is fresh and leaves the other tools unchanged.
Why did Grok name me last week and not this week?
Most likely something recent stopped being recent. A live feed weights what is being said now, so a mention that carried you into the answer stops carrying you once the conversation moves on.
Can I pay to appear in Grok answers?
There is no placement to buy. Paid attention on the feed can move a result while it is fresh, which is not the same as being recommended, and it stops the moment the spending does.
How do I find out what Grok already says about me?
Ask it the questions your buyers ask, in their words, and write down who gets named. The AnswerHalo free check does this for ChatGPT with 25 real buyer questions and emails the readout. It costs $0 and it runs as soon as you ask for it.
Does the paid audit cover Grok as well?
Yes. The $500 audit runs 125 checks across five AI tools, records who got cited on each one, and ends with the page that wins your worst question, specified section by section. It is in your inbox by this time tomorrow.
Your buyers are already asking. Find out what AI tells them.
The free check asks ChatGPT 25 real buyer questions about your niche and sends you the report within the hour, at $0. If it shows you are already getting named everywhere, we will say so in plain words.
Get my free check →